Vol. 5, No. 6, December 2012

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Volume 5. No. 5. December, 2012

December, 2012 www.ajocict.net

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Volume 5. No. 5. December, 2012

December, 2012 www.ajocict.net

A Journal of the Institute of Electrical & Electronics Engineers (IEEE) Computer Chapter Nigeria Section

ISSN- 2006-1781

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Published Papers

1-9 Approximations to Performance Metrics of Parallel Computer Systems Using Optimization and Computational Intelligence Techniques O.E. Oguike1; M.N. Agu & S.C. Echezona Department of Computer Science, University of Nigeria, Nsukka, Enugu State, Nigeria

10-19 Towards An Improved Stegano-Cryptographic Model For Secured Electronic Voting O.M. Olaniyi, O.T. Arulogun & E.O. Omidiora Department of Computer Engineering, Federal University of Technology, Minna , Nigeria.

20-26 Ray Tracing Characterization of Wideband Propagation Channel for Simulation of Mobile Radio Communications. A.C.O. Azubogu, C.O. Ohaneme, S.U. Ufoaroh and S.U. Nnebe Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria

27-33 Engendering Quality Information Systems Development through Software Architectural Design I. Gambo, P. Achimugu, R. Ikono, O. Iroju & A. Soriyan Computer Science and Engineering Department, Obafemi Awolowo University, Ile-Ife, Osun State

34-43 Mobility scenario-based Performance Evaluation of Preemptive DSR Protocol for MANET V.Ramesh & P.Subbaiah Research Scholar, Sathyabama University, Chennai, TN, India.

44-47 Development of a Mobile Feedback System for Health Institutions in Nigeria S. Okuboyejo, S. Akor & A. Adewumi Department of Computer and Information Sciences, Covenant University, Ota, Nigeria

48-54 Towards Developing an Online Social Media-based Mobile Learning System N.A. Ikhu-Omoregbe, C.K. Ayo, A.A. Azeta & V. Macus Covenant University, College of Science and Technology, Ota Nigeria

55-64 Enhancement of Console-based Dumpbin System to Display API Functions S.C. Chiemeke & O.E. Osaghae Department of Computer Science, University of Benin, Benin City, Nigeria.

65-72 Verifying and Validating Recursive Performance Models of Parallel Computer System Using Z-Transform O.E. Oguike, S.C.Echezona & M.N. Agu Department of Computer Science, University of Nigeria, Nsukka, Nigeria

73-80 Improved Electoral Fraud Prevention Mechanism for Efficient and Credible Elections in Nigeria F.O. Aranuwa & Oluwafemi Oriola Department of Computer Science, Adekunle Ajasin University, Akungba – Akoko, Ondo State,Nigeria

81-90 Modeling Variation of Waiting Time of Distributed Memory Heterogeneous Parallel Computer System Using Recursive Models O.E. Oguike, M.N. Agu & Echezona, S.C. Department of Computer Science, University of Nigeria, Nsukka, Nigeria.

91-97 Usable Authentication Schemes: A Critique O. O. Ayannuga Ph.D, and O. N. Lawal Computer Technology Department, Yaba College of Technology, Yaba, Lagos State, Nigeria.

98-103 Evaluating the impact of IT empowerment initiatives on income generation, employment and productivity. Monica N. Agu, Department of Computer Science, University of Nigeria, Nsukka, Nigeria.

iii

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

104-109 An Online Essay- Based Examination Assessment Model Using Double Blind Marking Technique C.O. Akanbi & Adetunji, A.B. Department of ICT ,Osun State University, Osogbo, Nigeria

110-115 Support Vector Machine for improving Performance of TCP on Hybrid Network A. Makolo Department of Computer Science, University of Ibadan, Ibadan, Nigeria

116-122 A Phone Learning Model for Enhancing Productivity of Visually Impaired Civil Servants A. A. Azeta & I.V. Azeta Department of Computer and Information Sciences, Covenant University, Ota, Nigeria

123-127 Towards Sustainable Energy Development Using Virtual Power Plants C.G. Monyei Department of Electrical and Electronic Engineering University of Ibadan, Ibadan, Nigeria

Call for Papers

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Editorial Board

Editor-in-Chief

Prof. Dele Oluwade Senior Member (IEEE) & Chair IEEE Nigeria – Computer Chapter. College of Information & Communication Technology Salem University, Lokoja, Nigeria

Editorial Advisory Board

Prof. Gloria Chukwudebe - Senior Member & Chairman IEEE Nigeria Section

Engr. Tunde Salihu – Senior Member & Former Chairman IEEE Nigeria Section

Prof. Adenike Osofisan - University of Ibadan, Nigeria

Prof. Amos David – Universite Nancy2, France

Prof. Clement K. Dzidonu – President Accra Institute of Technology, Ghana

Prof. Adebayo Adeyemi – Vice Chancellor, Bells University, Nigeria

Prof. S.C. Chiemeke – University of Benin, Nigeria

Prof. Akaro Ibrahim Mainoma – DVC (Admin) Nasarawa State University, Nigeria

Dr. Richard Boateng – University of Ghana, Ghana.

Prof. Lynette Kvassny – Pennsylvania State University, USA

Prof. C.K. Ayo – Covenant University, Nigeria

Dr. Williams Obiozor – Bloomsburg University of Pennsylvania, USA

Prof Enoh Tangjong – University of Beau, Cameroon

Prof. Sulayman Sowe, United Nations University Institute of Advanced Studies, Japan

Dr. John Effah, University of Ghana Business School, Ghana

Mr. Colin Thakur - Durban University of Technology, South Africa

Mr. Adegoke, M.A. – Bells University of Technology, Ota, Nigeria

Managing/Production Editor

Dr. Longe Olumide PhD Fulbright Fellow & Research Scholar Southern University and A & M College Baton Rouge, LA, USA

Universitry Academic Department of Computer Science University of Ibadan, Ibadan, Nigeria

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Foreward

The African Journal of Computing & ICT remains at the nexus of providing a platform for contributions to discourses, developments, growth and implementation of Computing and ICT initiatives by providing an avenue for scholars from the developing countries and other nations across the world to contribute to the solution paradigm through timely dissemination of research findings as well as new insights into how to identify and mitigate possible unintended consequences of ICTs. Published papers presented in this voume provide distinctive perspective on practical issues, opportunities and dimensions to the possibilities that ICTs offer the African Society and humanity at large. Of note are the increasing multi-disciplinary flavours now being demonstrated by authors collaborating to publish papers that reflect the beauty of synergistic academic and purpose-driven research. Obviously, these developments will drive growth and development in ICTs in Africa.

This issue of the African Journal of Computing & ICTs contains journal articles with a variety of perspective on theoretical and practical research conducted by well-grounded scholars within the sphere of information technology and allied fields across the globe. While welcoming you to peruse this volume of the African Journal of Computing and ICTs, we encourage you to submit your manuscript for consideration in future issues of the Journal

Have a fruitful reading

Thank you

Longe Olumide Babatope PhD Managing Editor Afr J Comp & ICTs

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Approximations to Performance Metrics of Parallel Computer Systems Using Optimization and Computational Intelligence Techniques

1O.E. Oguike, 2M.N. Agu & 3S.C. Echezona Department of Computer Science University of Nigeria Nsukka, Nigeria [email protected], [email protected]; [email protected] :

ABSTRACT Abstract— One of the reason for the use of a model as an approximation to another model is that the approximated model will be used as a substitute for the model that it is approximating. This will be very useful when the approximated model is more computational efficient than the model that it is approximating. Before a model can be used as an approximation to another model, the approximation conditions, if any, must be satisfied, this is very important because if the approximation conditions are not satisfied, the approximations will not be very effective. Therefore, in order to ensure that the approximations are very effective, the conditions for approximation must hold before the approximation model can be used. This paper uses optimization and nature inspired computational model to develop models that can be used as approximations to some of the performance metrics of heterogeneous parallel computer systems. The paper also states the various approximation conditions that must hold before such approximations can be very effective.

Keywords- Recursive models; performance modeling; approximations; parallel computer; distributed memory; parallel computer system; queuing network; variations; approximation condition..

1. INTRODUCTION

Optimization technique aims at producing efficient ways of doing things. Different optimization techniques are available for solving different problems with the aim of producing efficient results. Queuing systems is one of the optimization techniques can be used cpu queue to solve some problems that will model some of the performance Parallel metrics of heterogeneous distributed memory parallel computer Pocessors system. On the other hand, computational intelligence techniques cpu queue are nature inspired techniques that can be used to solve a particular problem. Recursion can be considered as computational intelligence technique because it is similar to natural echo, which continues to repeat itself, each time it repeats itself its intensity cpu queue diminishes until the natural echo stops. The optimization technique, queuing systems and the computational intelligence technique, recursion will be used to develop models that will be cpu queue used as approximations to some of the performance metrics of heterogeneous distributed memory parallel computer system. The effectiveness of the approximation depends on the approximation conditions. I/O queue

The parallel computer system under consideration is a heterogeneous distributed memory parallel computer system. It is I/O queue heterogeneous because the various computational resources, like processors and memory have different speeds and capacities, respectively, and the processes that are executing in parallel are I/O queue heterogeneous. The memory of the parallel computer system is distributed because each of the processors has its own memory. Since each of the processors has its memory, it implies that each of I/O processors the processors will have its own finite queue. Figure 1 shows a model of heterogeneous distributed memory parallel computer Figure 1: Queuing network of a heterogeneous parallel under consideration. computer system with distributed memory.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

We assume that the various queues are finite [1, 2, 3, 4] i.e. there is  , x  0,1,2,,3,...X  2 a limit to the number of jobs that can be admitted into the queues, i i i (1) x i   and negligible communication overhead. Suppose X1, X2, X3, … , 0, otherwise Xn , Xn+1, Xn+2, Xn+3, …, Xn+K are the maximum number of processes that can be admitted into the respective queues. We Since the various processors are heterogeneous, therefore, it assume that processes arrive at the various queues according to implies that the departure rate will vary, which can be described as: Poisson distribution, and they are serviced according to Exponential distribution [5, 6].  , x 1,2,3,4,..., X 1  i i (2)  xi   2. LITERATURE SURVEY 0, otherwise

Statistical models have been used as approximations to other statistical models provided that the approximation conditions hold. Using the steady state probability as stated in [7, 16] the The author in [33] pointed out that the Normal distribution models probability that x processes will be in the ith queue is could be used as approximations to Binomial and Poisson i distribution models, provided that the approximation conditions hold. In a similar manner, some statistical models can be used as x i P0i , x  X i 1 approximations to the performance metrics of parallel computer Pxi   (3) system, provided that the approximation conditions hold. The 0, otherwise authors in [29] developed analytic models that can be used to evaluate the performance of computer intensive applications of parallel computer system, while in [30], the authors developed The utilization factor for the ith queuing system,  i is defined analytic models that models the performance of compute intensive as: applications of a single processor computer system. The authors in [31] developed analytic models that can be used as approximations to the various performance metrics of J2EE application server. i . To obtain the value of P0i in equation (3), we sum all the Furthermore, the authors in [34] used fork/join station in a closed  queuing network with inputs from multi server stations to provide i efficient approximations to the performance measures. In a similar probabilities for the ith queue and equate it to 1. This implies that: manner, the authors in [35] used fork/join stations to obtain close approximation to the throughput of a close network with single X i 1 . (4) fork/join station that receives inputs from multi-station  Pxi  1 subnetworks. xi 0

3. DERIVING THE APPROXIMATIONS From equations (3) and (4), it implies that:

The approximation models are first derived for one queue, 2 3 4 X -1 P0i+iP0i+i P0i + i P0i + i P0i + … + i i P0i = 1. (5) afterwards, they are generalized for all the queues of the queuing network. Suppose Xi denotes the maximum number of processes Factorizing equation (5), it implies that that can be in the ith finite queuing system at any time [12, 13,], and 2 3 X -1  i denotes the utilization factor for the ith queue. P0i (1 + i + i + i + … + i i ) = 1. (6) A Models Based on one Queue. A computational intelligence model in form of recursive model can The following approximation models are based on one queue. be used to show where the series in (6) with Xi as the last term converges. The recursive model has been developed in [32] and it  Probability Density Function of the Number of Processes is given below as: in a Queue.

1 , X = 0 Let X denotes the maximum number of processes that can be on i i the ith finite queuing system at any time [12, 13]. Therefore, it means that the maximum number of processes that can be admitted Sum1(X , ) = (7) into the ith queue (excluding the process the processor is i  executing) is Xi-1. Suppose the arrival rate, xi when xi processes are in the ith queue of the queuing network be described as: Term1(X, ) + Sum1(X-1, ), X  0

Term1(X, ) is the recursive model that determines the xth term

of the series in (6), it is given as:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

1, Xi = 0  1  2 3 X i 2 (14) E(xi )   i 1  2i  3i  4i  ...  (X  1)i i   Sum1(X - 1,  )   i i  Term1i(Xi,  i ) = (8) A recursive model has been used in [30,31] to determine the convergence of the series in equation (14). The recursive model is i * Term1i(Xi-1,  i ), Xi  0 called Sum2i(Xi,  ), and it is given as: i Using equation (7) in equation (6), we obtain the following: 1, Xi = 1

Poi Sum1(Xi,  i ) = 1 (9) (15) Solving for Poi in equation (9), we obtain the following:

1 Term2 (X )*term1 (X -1, ) + Sum2 (X -1,  ), X 1 Poi = (10) i i i i  i i i i i

Sum1(Xi , i )

Therefore, summing the series with X-1 as the last term in order to It determines the convergence of the series in equation (14), with X as the last term. obtain the probability that x processes are in the queue, we use i Term2i(Xi) in equation (15) is given as: equation (10) in equation (3), to obtain the following: 1, Xi = 1

  xi  i , x  X 1 P  i (11) i x Sum1(Xi -1, i ) Term2i(Xi) = i  0,Otherwise (16)

1 + term2i(Xi-1), Xi  1 Equation (11) is the probability density function that models the term1i(Xi, ) is the recursive model in equation (8). probability that xi processes will be admitted in the ith queue. Therefore, using equation (15) in equation (14), and considering the  Approximation to Average Number of Processes in One series with X-1 as the last term, we obtain: Queue  Sum2(X -1,  )  Furthermore, the average number of processes in the ith queue Lq  E(x )   i i  (17) (i.e the queue, excluding the one being executed by the processor) i i    Sum1(Xi -1, i )  can be described statistically as expectation of xi , where xi is the random variable that denotes the number of processes in the ith Equation (17) can be used as approximation to the average number queue. This can be written as of processes in one queue, provided that the approximation conditions hold, which are large memory, i.e Xi must be large and

X i 1 the utilization factor for the ith queue must be greater than 1 and large enough. E( xi ) = xi Pi . (12)  xi x 0 i  Approximation to the Waiting Time in a Queue. Using Little’s formulae as stated in [7], the average length of Using equation (11) in equation (12), we obtain the following: the ith queue is directly proportional to the average waiting time in x X i 1  x  i  the ith queue. This can be expressed as follows,  i i  E(xi )   (13)  Sum1(X -1,  )  Lq  Wq (18) xi 1  i i  i i

Using equation (17) and the constant of proportionality in Equation (13) can be simplified as: equation (18), we obtain:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

 Sum2(X -1,  )  system. It is necessary to define i as the probability that a process W   i i  (19) i qi   Sum1(X -1,  )   e ffi i i  will join the ith queue after each cpu burst, and 0 as the probability that the execution of a process has been completed. Equation (19) is an approximation to the waiting time in the ith Arrival of processes into the various parallel processor queues can come from the outside world or from the various I/O queues or queue.  is the constant of proportionality, which is the e ffi from the particular parallel processor, at the expiration of the time effective arrival rate for the ith queue, which can be modelled as: quantum for that process. Let  be the rate of arrival of processes i effi = i(1-PX i -1) (20) into the ith queuing system, and  , the rate of arrival of processes from the outside world.. Under the steady state, when we consider Approximation to the Waiting Time in a Queuing System the queuing network, the overall utilization factor has been defined in [31] as: The waiting time in one queuing system considers the waiting   time in a queue together with the average service time of one  , i  0 processor. Using one of the queuing system laws, which says  0i that the waiting time in a queuing system is equal to the waiting i   (25)  time in a queue plus the average service time. Since the service  i , i  1,2,3,..., n  k has Exponential distribution, therefore, the waiting time in the    ith queuing system can be modeled as:  0 i Approximation to Number of Processes in all the Queues of the Queuing Network. 1 W  W  (21) si qi i Suppose xi is the random variable that denotes the number of Using equation (19) in equation (21), we have the following: processes in the ith queue. Therefore, another random variable, Y, can denote the total number of processes in all the queues of the  Sum2(X -1,  )  1 queuing network, as: W   i i   (22) si   Sum1(X -1,  )   nk  e ffi i i  i (26) Y   xi Approximation to the Average Number of Processes in One i1 Queuing System. Therefore, the average of the total number of processes in all the queues of the queuing network can be defined statistically as: Using Little’s formula, which states that the average number of processes in the ith queuing system is directly proportional to  nk  (27) the average waiting time in the ith queuing system. This can be E(Y)  E xi  expressed as:  i1  Using one of probability theory laws in [15,23], we obtain:

Lsi  Wsi (23) nk (28) Using the constant of proportionality, and equation (22) in E(Y)   E(xi ) equation (23), we have the following: i1

 Using equation (17) in equation (28), we obtain the following:  Sum2(Xi -1, i )  e ffi Ls     (24) i   nk  Sum1(Xi -1, i )  i   Sum2(X -1,  )   i i i  Lq  E(Y )    (29) i1 Sum1(X -1,  )  i i 

Equation (29) is the approximation to the average of the total number of processes in all the queues of the queuing network. B. Models Based on all the Queues..

Having developed the approximations to the performance metrics Approximation to Waiting Time in all the Queues of the of one queue, the approximations can be extended to all the queues Network of the queuing network of a heterogeneous parallel computer

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Using the extended Little’s formulae, which states that the Using the chosen constant of proportionality and equation (33) average number of processes in all the queues of the queuing in equation (34), we obtain the following: network is directly proportional to the average waiting time in all the queues of the queuing network, this implies,  nk      nk  Sum2(X -1,  )   e ff i nk  1  (35) Ws    i i    i1    Lq  Wq (30)  Sum1(X -1,  )   n  k     i1  i i    i1  i  Using the chosen constant of proportionality as stated in [31],   and using equation (29) in equation (30), we obtain the following: 4. SIMULATION RESULTS OF THE APPROXIMATIONS

The two approximation conditions were used to obtain the  nk  Sum2(X -1,  )    i i  simulation results of the approximations with the aim of showing   the effectiveness of the approximations when the approximation  i1  Sum1(Xi -1, i )  (31) Wq  conditions hold and weak or poor approximations, when the two  nk     approximations conditions do not hold [10]. Suppose the following  e ff i are the constant input parameters of a two-processor parallel  i1  computer: departure rates of processors 1 and 2 are 20 and 15,  n  k    respectively, probability of a process going to queue 1 and 2 are 0.5   and 0.3, respectively, and the probability that a process will leave the network is 0.2. The maximum number of processes in queue 1 Equation (31) is the approximation to the total waiting time in and queue 2 are 30 and 25, respectively. Table 1 and figure 2 show all the queues of the queuing network. the results of the waiting time approximation for the various values of the arrival rates. Approximation to the Waiting Time in all the Queuing System of the Network From the results in table 1 and figure 2, it can be seen that for Using one of the queuing system laws, which states that the various values of the arrival rates, such that the overall utilization total waiting time in all the queuing systems of the queuing factor for each of the queues is high and greater than 1 and for high network is equals to the total waiting time in all the queues of memory capacity, i.e. maximum number of processes that can be in the queuing network, together with the total of the average all the queues are large, the approximation for the total waiting time service time of all the processors. This can be written as: in all the queues is very effective, however, for various values of the arrival rates such that the overall utilization factor for each of the queues is low and less than 1, though the memory capacities of the nk  1  two queues are high, the approximation of the waiting time in all Ws  Wq    (32) the queues is not very effective. Furthermore, the results in figure 3 i1  i  and table 2 show the approximation results for the following constant values of input parameters of a two-processor parallel Using equation (31) in equation (32), we obtain the following: computer: departure rates of processors 1 and 2 are 20 and 15, respectively, probability of a process going to queue 1 and 2 are 0.5  nk  Sum2(X -1,  )    i i  and 0.3, respectively, and the probability that a process will leave   nk the network is 0.2.  i1  Sum1(Xi -1, i )   1  Ws     (33) nk   The maximum number of processes in queue 1 and queue 2 are 8   i1  i     and 5, respectively. It can be seen from the results in figure 3 and  e ff i  i1  table 2 that for various values of the arrival rates such that the  n  k  overall utilization factor is high and greater than 1, with low   memory capacities for the two processors, the approximation of the   waiting time in all the queuing system is not very effective, however, for low overall utilization factor that is less than 1 for Equation (33) is the approximation to the waiting time in all the each of the queues, and low memory capacities for the two queuing systems of the queuing network. Approximation to the processors, the approximation of the waiting time in all the queuing Number of Processes in all the Queuing System of the Network systems is poor. The reason for the ineffective and poor Using Little’s formula, which states that the number of approximations is because the two approximation conditions do not processes in all the queuing systems is directly proportional to hold. the waiting time in all the queuing systems. This can be written as: Ls  Ws (34)

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

With the same constant input parameters as in figure 2 and table 1, [6] Shanti Subramanyam, Performance Modelling of a J2EE we see in figure 4 and table 3 that for various values of the arrival Application to meet Service Level s, Agreement, Proc. rates such that the overall utilization factor for each of the queues is International Conference of Computer Measurement Group, (2005) high and greater than 1, the approximation of the number of processes in all the queuing system is very effective, but the [7] Hamdy A. T.,. Operation Research: An Introduction, Prentice- Hall of India, (1999). approximation is not very effective for various values of the arrival rates such that the overall utilization factor is low and less than 1. [9] Ivan Stojmenovic; Recursive Algorithms in Computer Science Courses : Fibonacci Numbers and Binomial The reason for this result is based on the fulfillment and non- Coefficients; IEEE Transactions on Education; Vol. 48, No. 3 fulfillment of the approximation conditions. [10] Arjan J.C. van Gemund; Performance Modelling of Parallel Systems: An Introduction. Similarly, the results in figures 5, 6 and 7 and in tables 4, 5 and 6 [11] Justyna Berlinska, The Statistical models of parallel show the results of approximated performance metrics and results applications, Annales UMCS Informatica, (2005). of the performance metrics. The effectiveness, poor and weak [12] Arranchenkov, K.E., Vilchersky, N.O., Shevlyakor, G.L approximations are as a result of the fact the two approximation Priority queueing with finite buffer size and randomized push- conditions hold or they do not hold. out; mechanism. Proc. of ACM SIGMETRICS international conference on measurement and modeling of computer 5. SUMMARY AND CONCLUSION systems.; (2003). [13] Abunday, B.D., and Khorram, E. The finite source queueing model for multiprogrammed computer systems with different This paper has been able to develop approximations to the various CPU times and different I/O times. Acta Cybern. 8, 4 , (1998) performance metrics of heterogeneous parallel computer system. [15] Trivedi K. Shridharbhai, Probability and Statistics with The approximations have been simulated on the computer together Reliability, Queuing and Computer Science Applications, with the actual performance metrics. The results of the simulations John Wiley & Sons Inc., (2002). have been analyzed to determine the effectiveness of the [16] Per Brinch Hansen. Operating System Principles. Prentice- approximations. The paper has discovered that the effectiveness of Hall of India Private Limited, (1990). the approximations depends on the two approximations conditions. [23] Robert V. Hogg and Allen T. Craig; Introduction to If the two approximation conditions hold, the approximations are Mathematical Statistics; Macmillan Publishing Co. Inc.; effective, if one holds and the other do not hold, the approximations (1978). are not effective, if the two approximations conditions do not hold [29] O.E. Oguike et al; Modelling the Performance of Computer the approximations are weak or poor. Therefore, provided that the Intensive Applications of Parallel Computer System; Proc. Of nd two approximations conditions hold, the paper has developed IEEE 2 International Conference on Computational alternative models, which serve as approximations to the various Intelligence, Modeling and Simulation; (2010). performance metrics of heterogeneous distributed memory parallel [30] O.E. Oguike et al; Performance Metrics of Compute Intensive computer system. Applications of Single Processor Computer System; Proc. of fifth Asian Modelling Symposium; (2011). Using an actual simulation package to simulate the performance of an actual heterogeneous parallel computer system can constitute an [31] Wei Xiong et al; An Analytical Approach for Performance Analysis of J2EE Application servers; extension to this research work, and comparing the results of the actual simulation with the results of the simulation models [32] O.E. Oguike et al; Evaluating the Performance of Heterogeneous Distributed Memory Parallel Computer developed in this paper. System Using Recursive Models; Proc. of Second International Conference on Intelligent Systems, Modelling REFERENCES and Simulation; 2011. [33] Keith Stoodley; Applied and Computational Statistics: a first [1] Henry H. Liu and Pat V. Crain, An Analytic Model for course; Ellis Horwood; 1984. Predicting the Performance of SOA-Based Enterprise [34] Nico Goossens et al; Approximations for Fork/Join Systems Software Applications, Proc. International Conference of with Inputs from Multi-Server Stations. Computer Measurement Group, (2004). [35] Erkut S¨onmez et al; An Analytical Approximation for the [2] S. Balsamo et al, A Review of Queueing Network Models Throughput of a Closed Fork/Join Network with Multi- with Finite Capacity Queues for Software Architecture Station Inpu Subnetworks Performance Prediction, (2002). [3] Catalina M. Liado et al, A Performance Model Web Service, Proc. International Conference of Computer Measurement Group, (2005). [4] Rosselio, J et al, A Web Service for Solving Queueing Network Models Using PMIF. www.perfeng.com/paperndx.htm, (2005). [5] Cathy H. Xia, Zhen Liu., Queueing systems with long-range dependent input process and subexponential service time. Proc. ACM SIGMETRICS international conference on Measurement and modeling of computer systems,(2003).

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

TABLE 1 APPROXIMATED WAITING TIME AGAINST ARRIVAL RATE FOR Approximated Waiting Time and Waiting Time HIGH MEMORY Against Arrival Rate

AR WT AWT 3.5 2 0.15 0.26 3 2.5 4 0.21 0.32

2 WT 6 0.37 0.49 1.5 AWT 8 1.21 1.29 1 10 2.24 2.27

0.5 Approximated Waiting Waiting Approximated Time and Waiting Time Waiting and Time 0 12 2.76 2.76 0 10 20 30 14 2.92 2.93 Arrival Rate 16 2.99 2.99

Figure 2: Approximated Waiting Time Against Arrival Rate 18 3.03 3.03 20 3.05 3.05 Approximated Waiting Time in all the Queuing TABLE 2 APPROXIMATED WAITING TIME AGAINST ARRIVAL RATE Systems and Waiting Time in all the Queuing OR LOW MEMORY System Against Arrival Rate AR WS AWS 0.6 2 0.029 0.145 0.5

0.4 4 0.087 0.2 WS 0.3 6 0.182 0.278 AWS Systems. 0.2 8 0.292 0.361 0.1

all the Queuing System and theQueuing System all 0

Approximated Waiting Time in Approximated Time Waiting 10 0.381 0.427 Waiting Time in all theQueuing all in Time Waiting 0 5 10 15 20 25 12 0.441 0.472 Arrival Rate 14 0.48 0.502

Figure 3: Approximated Waiting Time in all the queuing system Against 16 0.507 0.523 Arrival Rate 18 0.527 0.539 Length of all Queuing Systems and Approximated Length of all Queuing Systems 20 0.541 0.551 Against Arrival Rate TABLE 3 APPROXIMATED LENGTH OF THE QUEUING SYSTEMS AGAINST ARRIVAL RATE 60 50 AR LS ALS 40 LS 2 0.583 1.05 30 ALS 4 1.667 2.6

20 6 4.496 5.895 Systems and and Systems

QueuingSystems 10 8 18.921 20.232 Length of all Queuing Length of all

0 10 38.531 39.043 Approximated Length of all Length of all Approximated 0 10 20 30 12 48.229 48.305 Arrival Rate 14 51.171 51.214 16 52.333 52.375 Figure 4: Approximated Length of the Queuing Systems Against Arrival Rate 18 52.95 52.99 20 53.333 53.375

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

TABLE 4 APPROXIMATED LENGTH OF THE QUEUING SYSTEMS AGAINST ARRIVAL RATE

AR LS ALS

2 0.583 1.048

4 1.624 2.509 6 3.478 4.526

8 5.868 6.699 10 7.895 8.421

12 9.262 9.585

14 10.144 10.352 16 10.731 10.874

18 11.138 11.242 FIGURE 5: APPROXIMATED LENGTH OF THE QUEUING SYSTEMS AGAINST 20 11.431 11.512 ARRIVAL RATE

TABLE 5 APPROXIMATED LENGTH OF THE QUEUE AGAINST ARRIVAL RATE

AR LQ ALQ

2 0.117 0.583 4 0.733 1.667

6 3.096 4.495 8 17.093 18.405

10 36.523 37.037 12 46.189 46.265

14 49.129 49.172 16 50.292 50.334

18 50.908 50.95 FIGURE 6: APPROXIMATED LENGTH OF THE QUEUE AGAINST ARRIVAL 20 51.292 51.333 RATE

Table 6 APPROXIMATED LENGTH OF THE QUEUE Against Arrival rate

AR LQ ALQ 2 0.116 0.582 4 0.694 1.584 6 2.117 3.188 8 4.193 5.063

10 6.045 6.613 12 7.324 7.684

14 8.162 8.401 16 8.725 8.892 18 9.118 9.242

FIGURE 7: APPROXIMATED LENGTH OF THE QUEUE AGAINST ARRIVAL 20 9.404 9.5 RATE

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Author’s Brief Oguike, Osondu Everestus is a Senior Lecturer in the Department of Computer Science, University of Nigeria, Nsukka, Enugu

State, Nigeria. He has received many academic

prizes and scholarships as a result of his

outstanding academic performance. He is

interested in modeling the performance of parallel computer system. He can be reached by phone on +2348035405100 and through E-mail [email protected]

Dr (Mrs) Monica N. Agu is of Department of Computer Science, University of Nigeria, Nsukka, in the faculty of Physical Sciences. Her research has focused on using Information and Communication Technology on Poverty Alleviation and Modelling the performance of Computer Systems. She can be reached by phone on +2348039329480 and through E-mail [email protected]

Echezona Stephenson C. is a faculty member, as well as, on a PhD of the Computer Science Department, University of Nigeria, Nsukka. His research has focused on using Z-Transform to corroborate the Recursive Performance Models of Parallel Computer Systems. He can be reached by phone on +2348037325573 and e-mail through [email protected] and [email protected].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Towards an Improved Stegano-Cryptographic Model for Secured Electronic Voting

1Olaniyi, O.M, 2Arulogun O. T. and 3Omidiora E.O, 1Department of Computer Engineering Federal University of Technology, Minna , Nigeria. 2&3Department of Computer Science and Engineering Ladoke Akintola University of Technology, Ogbomoso, Nigeria. [email protected], [email protected] and [email protected]

ABSTRACT The widespread adoption of Information and Communication Technologies in governance over the years by electronic government has influenced democratic decision making through electronic voting (e-voting).E-voting provides increased participation of populace, reduced cost, decreased cases of invalid votes and support basis of democracy. However, e-voting systems are generally prone to security risks ranging from unauthorized casting of votes; Impersonation of voters by an attacker; Electronic ballot stuffing; Attack due to Denial of Service (DoS) and Distributed Denial of Service (DDOS) to the voting channel; Modification of vote and Deletion of valid votes. We present architecture for the development of an improved Stegano-Cryptographic model for electronic voting. The successful implementation and evaluation of the model will permit the government agency to know their degree of preparedness to organize free, fair and credible elections in future electronic democratic dispensation.

Keywords- E-voting System, Steganography, Cryptography, Biometrics, Security.

1. INTRODUCTION

The power of opinion expression capabilities and trustworthy Nowadays, the development and widespread use of Information and elections is the bedrock of democratic societies. Election, the Communication Technologies (ICT) is linked to the proper gateway of confidential democratic governance, is the fundamental execution of democratic rights and thus, ICT is changing democratic instrument and most acceptable means around which people convey decision making through electronic democracy(e-democracy).E- their views to their government by democratic process [1].The democracy has become a necessity in the era of Computing and democratic process rests on a fair, universally accessible voting Information Technology. Electronic voting (E- voting) as one of the system through which all citizens can easily and accurately cast a paramount pillars of e-democracy is the use of computerized voting vote [4]. Voting, an indispensable feature of democracy is a method equipment to cast and tabulate ballots in a trustable manner [6]. by which a group of people express their opinion over who will lead Electronic systems are used to register voters, count ballots and them for a specific period of time via electoral process. Usually record votes [38].E-voting can be identified in six ways [15]: Poll correctness, robustness to fraudulent behaviors, coherence, site Direct Recording Electronic (DRE) voting, remote Internet consistency, security and transparency of voting are all key based voting, Optical Scanning Systems, Voter Verified Audit Trail requirements for the integrity of an election process [24]. (VVAT), Voting Kiosk, and Non-Internet Remote Voting.

Traditionally, the process of voting is not only cumbersome as the Remote electronic system of voting offers multiple advantages voters are expected to vote in person, the cost and process of manual compare to traditional paper-based voting with the voting are increasing geometrically and tedious to execute [17]. This following[41;24;36] :1)Increased participation in democratic has resulted into declining participation rate due to inconvenience of governance as more citizens have access to express their current classical/manual system of voting like inaccuracy in ballot opinion;2)Reduced costs as the materials required for printing and counting and delayed announcement of election results [18]; Loss of distributing ballots as well as the manpower required to govern poll significant time during ballot counting [3]; Unacceptable sites are considerably reduced;3)Flexible as it can be tailored to percentages of lost, stolen and miscounted ballot papers, votes loss support multiple languages and permit up-to-date minute ballot through unclear or invalid ballot marks and limited accommodations modifications.4)Greater speed and accuracy in placing and tallying for people with disabilities [2]; Under-age voting, counting error, votes as e-voting step by step processes help minimize the number complicity of the security agencies and absence or late arrival of of miscast and rejected votes;5)Lower election fraud in endangered election materials[24][36]. countries with young democracies;6)Deliver voting results reliably and more quickly.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

According to literature, the design of secure voting system must Although, voters turnout will increase based on this proposed satisfy a number of competing criteria [35][36][34][19]. These framework, security of vote over the wireless channel is not requirements give an avenue for a free, fair, credible and guaranteed and as a result the integrity of votes can be confidential election. These requirements by [36] are grouped into compromised. Similar reference framework was proposed in [41] generic and system specific; by [19] as functional and non- with the view that secure voting protocols can only fulfill their functional requirements. Considering e-voting from generic point of objectives to electronic voting systems with the cooperation of view, the following requirements are necessary: a) Security: Votes voting authorities termed as Organizations; Data in the form of should not be manipulated during the whole process of voting; b) digital certificates and electronic ballot paper; Functions in the Confidentiality: No one should access any information about the name of the core algorithms for vote encoding and decoding, voter’s vote so as not to be able to alter it; Privacy: No one should signature algorithms and anonymous channels of transmission and be able to link the voter to this vote after casting a vote. c.) Digital Infrastructure in form of Computer Hardware and Authenticity: Only eligible voters can cast their votes. d.) Software. Design tasks and security analysis of electronic voting Integrity/accuracy: Votes cast cannot be altered by an attacker. All systems have to take account these important elements. valid votes must be counted, whereas all invalid votes must not be discarded; e) Convenience: Voters should be able to cast votes According to [40], there is a great need for e-voting and security in quickly with minimal equipment or skills; f.)Democracy: Permits e-voting systems. There is necessity to implement an additional only eligible voters to vote only once; g) Verifiability: Voting layer of security technology to tackle the risks posed by electronic systems should be verified so as to have confidence that they meet voting and ensure security requirements such as voters’ privacy necessary criteria. and vote integrity. Different mechanisms to ensure security of voting systems such as: Personal Identification Number or In other for above ideals of genuine election to provide confidence password, encryption, digital signature, smart cards and biometric and enable a peaceful resolution of the struggle for political power identifiers were proposed. The design of secure electronic voting between the leaders and followers in democratic governance; all system according to literature must satisfy a number of security aspects of elections process must be directly observable by the competing criteria [35][36][34][19]. candidates, the official observers and the people themselves. For people to be directly observable; transparency, integrity of electoral These security requirements give an avenue for a free, fair, credible process, security of lives and the process of elections must be fair and confidential election. Existing secure models for e-voting and guaranteed. Therefore, for e-voting system too bridge gap relies on the technology of cryptography by encryption. These created by classical traditional method of conducting genuine models can be classified into authoritative and non-authoritative elections, a list of security requirements that constitutes a must for models [28]. The non authoritative model like [13] is fewer while voting must be observed. Without these requirements, rigging, fraud the authoritative models can be categorized by different and corruption in electoral process will occur. These fundamental technologies into three schemes. These schemes are Homomorphic requirements include: confidentiality, integrity, authentication and scheme such as [5,30,31,10]; the blind signature such as verifiability/non-repudiation [1][18]. [9,43,7,16,50]; and the mix net scheme such as [32,27,45] .

This paper presents the application of steganography and However, these encryption algorithms have not only been proved cryptography to the design and development of an improved unreliable as computing power keeps increasing [42][44], they have stegano-cryptographically model for secure electronic voting. The been crypt- analytically found to be vulnerable to attacks ranging model is proposed for secure remote electronic voting system with from brute force attack, timing attack, session hijacking replay the view of increasing participation, confidence and trustworthiness attack , cipher-text-only, known-plaintext, chosen-plain text attack in electronic democracy, protects voter’s against intimidation, a, chosen-plain text and trapdoor problem[23][21]. Moreover usage provide sufficient evidence to convince the electorate to vote, of cryptography alone for the transmission of votes over insecure convince the losing candidate that he actually lost as a result of wireless medium through remote internet voting would undoubtedly conducted, free, fair, credible and genuine elections. threaten the integrity of democratic elections as attention of adversaries are drawn to access and attack the data being 2. RELATED WORKS transmitted. Since secure e-voting system must embrace secrecy of vote, combining the merit of multimedia data security obtainable There are quite a few literatures which exist in the area of security from essentials features of encryption schemes by cryptography with in electronic voting systems. The framework for mobile steganography can play a major role in shielding the vote casted by multilingual e-voting system on which the security measure in [21] remote voters in cyberspace to secure the privacy, confidentiality and [40] may be built was proposed in [37]. Due to the problems of and integrity of electronic voting. the existing manual voting system in most developing nations, this framework was introduced. The multilingual framework enables citizen to choose the language they best understand which will increase the turnout of voters as communication barriers will be broken.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Related works in literature in application of stegano-cryptographic The author’s proposed system provided a secure voting mechanism modelling from generic point of view to security issue electronic to the basic requirements of a secure voting system as well as non- voting domain exist. In [26], both techniques were integrated into a functional requirements like uncoercibility, receipt-freeness and multi-layer data security model of secret communication. The model universal verifiability by experimentation with two different cryptographic process was carried out using symmetric block steganographic tools, F5 and Outguess on five different types of ciphers with linear algebraic equation for message encryption while images. The results of the experimentation show that slight changes the steganographic process embedded the block cipher text obtained exist between original images and stego images after secret message from cryptographic process into the cover image to produce stego is embedded. Our proposition is premised along the exploration of image using Least Significant Bit (LSB) Image domain, Image multi layer: steganographic and cryptographic techniques of data Steganography. The result of simulations of the resultant bitmap security and multimedia: Image and video approach to the problem based Stego-image using Matlab shows a promising result with a of authentication, integrity, confidentiality, non-repudiation to the significant difference of the Peak Signal to Noise Ratio (PSNR) and problem of electronic voting for electronic democracy in developing Signal to Noise Ratio (SNR) of the stego image and the original countries. image. 3. CRYPTOGRAPHY AND CRYPTOGRAPHIC MODELS However, [26] steganographic process was carried out in image FOR E-VOTING SYSTEMS domain using LSB steganography on bitmap format (the Leena Image) which have low robustness against statistical attack from Cryptography is the science of securing data communication in the statistical steganalyst and low robustness against image presence of an adversary. By cryptography sensitive information can manipulation which might destroy the hidden message from its be stored and transmitted across insecure networks in manner such destination[29].Also, LSB steganography on bitmap format is that unintended antagonist cannot read the information except the generally suitable for application where focus is only on the amount intended recipient [39]. Cryptographic objectives encompass using of information to be transmitted and not on the secrecy of the mathematical techniques to all aspect of Information security from transmitted information due to above deficiency defeating the confidentiality, entity authentication and origin authentication and efficiency of the [26] crypto –steganographic model. data integrity [14]. Usually Party A, the sender, sends secret Biometric online voting scheme based on Stegano-cryptographic message to Party B, the intended receiver, over a communication modeling technique was proposed to the problem of authentication line which may be tapped by an adversary [17]. requirement of a voting system [48]. The scheme uses the principle of LSB based Image steganography as cover object and secret key 3.1 Components of Cryptography generated through cryptographic hash function for the scheme Cryptography encompasses many problems including cryptography. The scheme is based in the assumption that the authentication, encryption, key distribution, and decryption. The voter’s biometric fingerprint information, personal Identification traditional solution to these problems achieved through Private key Number and account creation of the vote are securely generated, Encryption (PKE). PKE involves the meeting and agreement of collected and available online for election. The performance of the Party A and Party B on a pair of encryption and decryption algorithm was analyzed and the result revealed that the scheme does algorithms Ɛ and D as well as common secret S, called Key, prior to not give any chance to steganalytic tools to search and predict set of remote transmission of sensitive information. The adversary may modification of attack. In [49], a secured electronic voting system have the knowledge of Ɛ and D but does not know S. After the prior was proposed using the Stegano-Cryptographic modeling technique. meeting, Party A encrypts message m by computing the ciphertext The system was implemented around the principles of secret ballot c=Ɛ(S,m) and sends c to B. Upon reception of an encrypted message theory, image steganography, visual cryptography and threshold c, Party B decrypts c by computing m=D(S,c).The adversary who decryption cryptosystems in Java. does not know S should not be able to determine message m from cipher text c [38;48].This is illustrated in the Figure 1

Fig 1: Symmetric-Key Cryptographic model (Source [14])

Vol 5. No. 6. Dec 2012 ISSN 2006- 1781 African Journal of Computing & ICT

In secure e-voting domain, after first cryptographic models voting, efficient tallying and large scale election support for electronic elections was published [10][13][5] several were carried out. Findings of comparisons in [34][1] was schemes have been proposed in literature to deal with the that blind signature model is the most efficient security problems in electronic voting. In [34][1], four cryptographic model for secure electronic voting as it proposed generic cryptographic models for secure electronic supports more core properties desirable for secure e-voting. voting were compared amongst their core properties of General framework of Cryptographic model to secure universal verifiability, support for write-in ballot, efficient electronic voting system is shown in Figure 2:

Figure 2: General Framework Cryptographic Model to Secure E-voting

3.2 STEGANOGRAPHY AND STEGANOGRAPHIC MODELS FOR E-VOTING SYSTEMS Steganography is the science of concealing digital information The modern formulation of steganography to an application within electronic files like image, sound, an article, a area is given in terms of prisoners’ problem in which Alice shopping list such that no-one determines that the hidden (the sender) and Bob (the receiver), the two inmates wishes to communication is taking place [30][22]. The technique secure communicate to formulate an escape plan without the data by obscuring and embedding the content in another knowledge of wendy, the prison warden. Supposing Alice media called carrier in which the information is saved for sends secret message M to Bob using steganophic process, he transmission. The technique of data security by simple chooses a cover medium which can be image, video and audio encryption is not sufficient anymore as technology of Super C. The steganographic algorithm employed as shown in figure Information Highway evolves. An encrypted data could easily 3 identifies C’s redundant bit and embed it to a chosen media, be suspicious.[49].Compare to cryptography which focuses on for instance image, to create a Stego Image, S, by replacing keeping the contents of message secret by scrambling these redundant bit with data from Message M. The Stego messages so it cannot be understood though its existence may image S is transmitted over insecure wireless link under the be detected, Steganography focuses on keeping the existence monitoring of Wendy to the receiver Bob only if Wendy has of message secret by striving to hide the presence of the no suspicion of it [14]. The process of embedding the message itself from an observer so there is no knowledge of privilege data for transmission in public channel represents a the existence of the message in the first place [22].Both critical task for steganographic system because the stego technologies can be combined to produce better protection of Image S must be as similar as possible to the chosen cover the information [46]. media for the avoidance of the eavesdropper.

Figure 3: General Steganographic Model (Source from [14])

Vol 5. No. 6. Dec 2012 ISSN 2006- 1781 African Journal of Computing & ICT

In secure e-voting domain, the steganographic message Some of the techniques used in steganography are domain consists of the secret message, the electronic ballot, the cover tools such as Least Significant Bit (LSB) insertion, noise data and the stego message. The secret message is the part of manipulation, and transform domain that involve the message intended to be hidden, the cover data refers to manipulation algorithms and image transformation such as the container for hiding the secret message and the stego discrete cosine transformation and wavelet transformation. message is the final product of steganography. The general However there are techniques that share the characteristic of framework for steganography to electronic voting is shown in both of the image and domain tools such as patchwork, Figure 4: pattern block encoding, spread spectrum methods and masking [49].

Figure 4: General framework of Steganography to E-voting (Adapted from [22])

4.0 STEGANO-CRYPTOGRAPHIC MODELLING TECHNIQUE

Stegano-Cryptographic modeling technique involves the combination of the two principal Information Security Technologies- Cryptography and Steganography to the problem of security in an application area. Considering Figure 3 and Figure 4, features peculiar to both information security techniques unify into this model called Stegano-cryptographic model or stego-cryptographic model shown in Figure 5. This new relationship exists as result of mapping between the plaintext P and Message M, Cipher Text E and Stego Media S and Cryptographic Key K and the Stego Key K.

Figure 5: General Stegano-Cryptographic Model Mapping from Steganography and Cryptography (Adapted from [14])

Vol 5. No. 6. Dec 2012 ISSN 2006- 1781 African Journal of Computing & ICT

The stegano-cryptographic model results as a hybrid model simultaneously. At the receiver side, Bob will be able to with the addition of a new element: the Stego key K, giving recover Secret Message M through Stego Media S and Stego the unifying model the cryptographic functionality while K. In addition, Wendy will neither detect that Secret preserving the desired steganographic attributes. The hybrid Messsage M is embedded in Stego Media S nor be able to model embedding process yields Stego Media S exploiting access the content of the secret message M [14].Figure 6 not only Cover Media C’s bits but also K’s ones as shown shows a classical example for an image based Stegano- in Figure 6.Therefore by Figure 6, Alice (the sender) will Cryptographic Model: have the privilege to embed the secret message M (that is, the plaintext) into the Cover media C (through steganographic process) encrypting Message M by the Cryptographic key K (Through cryptographic process)

Figure 6: General Framework of Image Based Stegano- Cryptographic model (Source [14])

In secure electronic voting application, the voter’s The secret key is used to extract the hidden message from electronic intent, vote is first encrypted using an encryption the stego media using the decryption algorithm. For instance algorithm. The encrypted message is then embedded into a in image steganographic application, the integrity of a voter stego media which can be image, video and audio and his vote is assured with the encryption of the message depending on the steganographic technique using a stego- (vote) and then embedding of the encrypted message inside key. The stego media is then sent through a communication a 24-bit cover image. A secret key used for the stego-system channel. encoder is then passed through the communication channel. At the voters administrator end, the secret key is used to extract the hidden message from the stego-image as shown in Figure 7.

Figure 7: Stegano-Cryptographic Modeling Technique in Secure E-voting.

Vol 5. No. 6. Dec 2012 ISSN 2006- 1781 African Journal of Computing & ICT

5. RESEARCH DIRECTION To achieve this, a study of the underlying principles of various cryptographic models for e-voting like Blind In this research, our aim is to develop a robust multi layer Signature, Homomorphic, Mix-net and verifiable secret (steganography and cryptography) data security, multi sharing ; various underlying Cryptosystems like RSA domain (Image, video and/ or audio) model to the problem Cipher and ECC Cipher and various techniques of image of fundamental security issues of authentication, integrity, and voice steganography will be carried out. The confidentiality and non-repudiation requirements of secured performance analysis of the final multi-layer, multi domain electronic voting system. model will be carried on an electronic voting system to reveal its vulnerability to fundamental security issues of authentication, integrity, confidentiality and non-repudiation requirements in secured e-voting system. The architecture of our proposed model is shown in figure 8:

PRE-ELECTION PHASE:

Laptop

Message1

FRONT END

REGISTRATION

ELECTION PHASE:

Authenticate User using Stegano-cryptography LOGIN Vote Biometrics Agorithm

POST-ELECTION PHASE: BACK END Database Stegano- DISPLAY LOGIN AS RETRIEVE cryptography COUNT FINAL ADMINISTRATOR RESULT Algorithm RESULT

Figure 8: Architecture of the proposed model of Secured E-Voting System

The architecture chosen for the proposed model is client--server The model will evaluate security issues on electronic voting and architecture based on three-tier architecture. A three-tier is a predict the performance of stegano-cryptographic secured client–server architecture in which the presentation, the electronic voting system for future electronic democracy. Our application processing, and the data management are logically research agenda is premised towards the development of a robust separate processes. The model will be simulated using Java security model for electronic voting using principles of Programming Language and Oracle Database Management Biometrics, Steganography and Cryptography. The improved System. The model will be evaluated to verify and ascertain its stegano-cryptographically model will verify voters as who they stability against established fundamental security requirements of claimed they are, prevent fraud in form of addition and deletion of secured e-voting system. ballot (vote) over an insecure wireless networks, protect voters privacy, ensures confidentiality and uncoercibility which are 6. CONCLUSION essential to security requirements of remote electronic voting systems. The outcome will be a benchmark for relevant Proper administration of elections is central to democracy. The government electoral agency (like INEC in Nigeria) to know their electronic administration of elections by electronic voting must degree of preparedness to envisage performance metrics require to provide a list of security requirements for fair and transparent organize free, fair and credible elections in future democratic democratic decision making. Without these requirements, rigging, dispensation. At this stage, the research is open to suggestions and fraud and corruption in electoral process will occur. The proposed criticisms. architecture of an improved stegano-cryptographic secured model for e-voting turns attention to data hiding techniques in steganography and cryptography in building stronger scheme that combines the strength of both information security and privacy technologies.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

REFERENCES [14]Domenico B. and Luca I. (2007), “Image Based Steganography and Cryptography” Institute for Systems and [1]Abo-Rizka M and Ghounam H.R(2007), A Novel in E-voting Technologies of Information, Control and Communication , pp in Egypt, International Journal of Computer Science and Network 127-134. Security ,Vol.7,No.11,pp226-234. [15]Enguehard C (2008),Transparency in Electronic Voting: The [2]Ayannuga O.O and Folorunso O (2010), Electronic Voter’s Great Challenge, Proceeding of International Political Authentication Management System (eVams), Proceedings of 23rd Association(IPSA) Conference on E-democracy-State of the art National Conference of Nigeria Computer Society July 26th -30 th and future agenda, Stellenbosch University, South Africa,22-28 2010,Volume 21,pp105-110. January,2008.

[3]Akinyede R.O (2010), Nigerian Voting System: Present and [16]Fujioka A., Okamoto T., and Ohta K.(1992), “A Practical Future States, Proceedings of 23rd National Conference of Nigeria Secret Voting Scheme for Large Scale Elections. In Advances in Computer Society July 26th -30 th, Volume 21,pp77-81. Cryptology , AUSCRYPT '92, pp. 244-251.

[4]Bannet J, Price D.W, Rudys A, Singer J and Wallach [17]Goldwasser S and Bellare M(2001),”Lecture Notes on S.D(2004), Hack-a-vote: Security Issues with Electronic Voting Cryptography”, MIT, Cambridge. Systems, IEEE Security and Privacy-IEEE Computer Society available at http://www.computer.org/security/ [18]Ibrahim S, Kamat M, Salleh M, and Abdul Aziz S (2003), Secure voting using blind signature available at URL [5]Benaloh, J. (1987),”Verifiable Secret Elections”, PhD http://eprints.utm.my/3262/1/IEEE02- Thesis,Yale University, New Haven. EVS_full_paper_ver14Nov.pdf Retrieved on November17th 2011. [6]Cetinkaya ,O and Koc, M,L(2009),Practical Aspects of [19]Kalaichevi V and Chandrasekaran R.M (2011),Secured Single DynaVote E-voting Protocol,Electronic Journal of E-Government, Transcation E-Voting Protocol: Design and Implementation, Volume 7Issue 4,pp327-338. European Journal of Scientific Research,Vol 51 No2,pp276- 284. [7]Cranor, L.R. and Cytron, R.K. (1996) “Design and [20] Lambrinoudakis C, Gritzallis D, Tsoumas V, Karyda M and Implementation of a Practical Security-Conscious Electronic Ikonomopulos (2003), Secure- Electronic Voting : The Current Polling System,”, Washington University: Computer Science Landscape, Advances in Information Security, Volume 7,No Technical Report. 2,pp 101-122 Available at http://www.springerlink.com/content/j4402372h6256372.pdf [8] Ciprian Stănică-Ezeanu (2008), “E-Voting Security”, Buletinul Universităţii Petrol – Gaze din Ploieşti, Vol. LX (2), pp 93-97. [21] Longe O.B., Boateng R., Dada E.G., Olaniyan O. and Olaseni O. (2010), “Stegacrypt: A Reduced Least Significant Bit [9] Chung-Ta L and Min Shang H (2012),” Security Enhancement Insertion Rate Carrier for Transmitting Embedded of Chang-Lee Anonymous E-voting Scheme”, International Information”, Journal of Computer Science and Its Journal of Smart Home,Vol.6 No2,pp45-52. Applications, Vol. 17(1), pp 1 – 11.

[10]Chaum D. (1981), “Untraceable electronic mail return [22]Longe, O.B. (2011c). On the use of Image-based Spam Mails addresses and digital psedonymns”, Communications of the ACM, as Carriers for Covert Data Transmission. Computing and Vol 24(2) ,pp84-86. Information Systems Journal, Vol. 15. Issue 1., Pp1-5.

[11]Chaum D. (1983),”Blind Signatures for untraceable [23]Longe O.B, Roberts A.B.C, Onifade O.F.W, Kaka O and payments”, In Proceedings of Cryptology, pp:199-203, Plenum Isiaka R.M (2008a),Framework for the development of a Press, NewYork. Hybrid Chaotic Image Scheme for Multimedia Data Encryption,3rd International Conference on ICT Applications, [12] Choonsik Park, Kazutomo Itoh, and Kaoru Application of ICT to Teaching, Research, and Administration Kurosawa(1993),“Efficient anonymous channel and all/nothing (AICTTRA 2008), Volume III, pp150-154, 21st -25th election scheme,” In Proceeding of Advances Cryptology - September 2011, Obafemi Awolowo University,Ile-Ife, Osun EUROCRYPT’93, pp.248–259, 1993. State, Nigeria.

[13 ]DeMillo, R. A., Lynch, N. A. M. Merritt,(1982) [24] Manish K, Suresh K.T, Hanumanthappa. M, Evangelin “Cryptographic Protocols,” In Proceeding of 14th Annual ACM G.D,(2005), Secure Mobile Based Voting System, Retrieved Symposium on Theory of Computing, pp.383-400. online at http:// www.iceg.net/2008/books/2/35_324_350.pdf on November17th 2011.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[25]Malkawi M, Khasaweh M K., and Al-Jarrah O, (2009), [37] Olaniyi, O.M, Adewumi D.O, Oluwatosin E.A, Arulogun, O. T ”Modelling and Simulation of a Robust E-voting System”, and Bashorun M.A (2011),” Framework for Multilingual Communication of Information Management Mobile E-Voting Service Infrastructure for Democratic Association(IBIMA) Journal, Volume 8, pp 198-206. Governance” .African Journal of Computing and ICTs, Vol. [26]Mallick P K and Kamilla (2011),Crypto Steganography Using 4, No. 3. Issue 2, pp 23 – 32. linear Equation, International Journal of Computer and Communication Technology, Volume 2 Issue8,pp106-112. [38] Olaniyan O.M, Mapayi T, Adejumo S.A (2011), “A Proposed Multiple Scan Biometric-Based System for Electronic Voting” [27]Markus M , and Patrick H,(1996) “Some remarks on a receipt- , African Journal of Computing and ICT September 2011, Vol. free and universally verifiable mix-type voting scheme,” In 4(2), pp 9 – 16. Proceedings of ASIACRYPT ’96, LNCS 1163,pp.125– 132,1996. [39] PGP Corporation (2003),”An Introduction to Cryptography”, PGP Corporation, USA. [28]Meng B (2009), “A Secure Internet Voting Protocol Based on Non Interactive Deniable Authentication Protocol and Proof [40] Sonja H. (2004), “E-Voting and Biometric Systems?” protocol that two Cipher Texts are Encryption of the Same Proceedings of the first international workshop, “Electronic Text”, Journal Of Networks, Vol. 4(5), pp370-377. Voting in Europe: Technology, Law, Politics and Society”, 2nd- 4th July 2004, Beautiful Castle Hofen, Europe, pp 63-72. [29]Morkel T,Eloff J.H.P and Olivier M.S(2010),”An overview of Image steganography”, Department of Computer [41] Schryen G, (2004),”Security Aspects of Internet Voting”, Science,University of Pretoria,South Africa.Retrived online at Proceeding of 37th Annual Hawaii International Conference on http http://martinolivier.com/open/stegoverview.pdf on 4th System Sciences(HICSS ‘04),Volume 5,pp. 50-61. June 2012 [30]Popa R, (1998) ,”An Analysis of Steganographic System”, The [42] Si H and Li C(2005),Maintaining Information Security in E- "Politechnica" University of Timisoara, Faculty of Automatics Government through Steganology, available at URLwww.igi- and Computers, Department of Computer Science and global.com/chapter/encyclopedia-digital- Software Engineering. government/11652.pdf [31] Ronald C, Matthew K.,Franklin, Berry S, and Moti Y(1996), Multi-authority Secret-ballot elections with linear work, In [43]Sujata M and Banshidhar M(2010),”A Secure Multi authority Advances in Cryptology, EUROCRYPT '96, v ol.1070 of Electronic Voting Protocol based on Blind LNCS, pp.72-83.Springer-Verlag, Signature”,Proceeding of the IEEE International Conference on Advances in Computer Engineering,pp 271-273. [32]Ronald C, Rosario G, and Berry Schoenmakers (1997), ” A Secure and optimally efficient multi-authority election Scheme”. In advances in Cryptology-Eurocrypt 97,pp 103- [44] Wang, X, Feng, D,Lai, X, and Yu H(2004).Collisions for Hash 118,Springer Verlag,LNCS. Functions MD4,MD5,HAVAL-128 and RIPEMD. Cryptology ePrint Archive, Report [33] Kazue S and Joe K,(1995),“Receipt-free mix-type voting 2004/199.http://eprint.iacr.org/2004/199.pdf scheme,” In Proceeding of EUROCRYPT ’95, LNCS 921,pp.393–403 [45]Emmanouil M, Mike B, and Vassilios C(2001), “Receipt- freeness in large-scale elections without untappable channels,” [34]NSF (2001),” Report on the National Workshop on Internet In Proceeding of I3E,pp.683–694,2001. Voting: Issues and Research Agenda” , National Science Foundation, Retrieved at [46]Katiyar S,Meka K R,Barbuiya F A,and Nandi S(2011),”Online http://news.findlaw.com/cnn/docs/voting/nsfe-voterprt.pdf. Voting System Powered by Biometric Security Using Steganography”, Proceedings of The Second International Conference on Emerging Applications of Information [35] Okediran O. O, Omidiora E.O, Olabiyisi S.O, and Ganiyu R Tehnology,IEEE Computer Society,pp 288-291. A (2011),” A Comparative Study of Generic Cryptographic models for Secure Electronic Voting”, British Journals of [47]Rura L,Isaac B, and Haldar M K, (2011),Secure Electronic Science Vol1No2, pp 135-142. Voting System Based on Image Steganography,Proceedings of IEEE Conference on Open Systems(ICOS2011), [36] Okediran O. O, Omidiora E.O, Olabiyisi S.O, Ganiyu R A and IEEE,September 25-28,2011,Langwi,Malaysia. Alo OO(2011),” A framework for a Multifaceted Electronic Voting System, International Journal of Applied [48]Hoffstein J,Pipher J and Sivermann J(2008),”An Introduction to Sciences,USA,Vol1 No4, pp 135-142. Mathematical Cryptography”, Springer,USA.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[49]Muhalim M A, Subariah I, Mazleena S and Mohd R K Authors’ Brief (2003),”Information HidingUsing Steagnography”,Faculty of Computer System and Information System, Department of Computer Science and Communication ,Universiti Technologi Olaniyi O. Mikail is a Lecturer in the Malaysia Department of Computer Engineering, Federal University of Technology, and [50] Wen-Sheng J, Chin-Laung L, and Pei-Ling Y,(2002), “A Minna, Niger State. He had his First Degree verifiable multi-authorities secret elections allowing in Computer Engineering at the Ladoke abstaining from voting”, Computer Journal 45(6),pp.672-682. Akintola University of Technology, Ogbomosho, Oyo State, Nigeria and M.Sc. Degree in Electronic and Computer Engineering at Lagos State University, Lagos. He is currently a doctoral student at the Department of Computer Science and Engineering, Ladoke Akintola University of Technology, Ogbomosho, Oyo State, Nigeria. He is a member of International Association of Engineers and Computer Scientists and Registered with the Council of Regulation of Engineering in Nigeria. He has published in reputable journals and learned conferences. His areas of research includes: Intelligent Systems, Computer Security, E- Governance, and Telemedicine. He can be reached on the cyberspace at [email protected] & Http://www.olaniyimikail.webs.com.

Arulogun O. T. is a Senior Lecturer in the Department of Computer Science and Engineering, Ladoke Akintola University of Technology, Ogbomoso, Nigeria. He is currently a visiting scholar at Hasso-Plattner Institute, Potsdam, Germany. He has published in reputable journals and learned conferences. His research interests include networks security, mobile IPv6, wireless sensor network and its applications. He belongs to the following Professional bodies: Computer Professionals (Registration) Council of Nigeria; Registered Engineer, COREN and International Electrical/Electronic Engineers (IEEE). He can be reached on the cyberspace at [email protected]

Omidiora E. O. is currently a Reader in the Department of Computer Science and Engineering, Ladoke Akintola University of Technology, Ogbomoso, Nigeria. He graduated with B.Sc. Computer Engineering from Obafemi Awolowo University, Ile-Ife, Nigeria. He bagged M.Sc. Computer Science from University of Lagos, Nigeria and Ph.D Computer Science from Ladoke Akintola University of Technology, Ogbomoso, Nigeria. He has published in reputable journals and learned conferences. His research interests include: The study of Biometric Systems, Computational Complexity measures and Soft Computing. He belongs to the following professional bodies: Full Member, Computer Professionals (Registration) Council of Nigeria; Corporate Member, Nigeria Society of Engineers; Register Engineer, COREN and . His contact email addresses are [email protected] and [email protected]

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Ray Tracing Characterization of Wideband Propagation Channel for Simulation of Mobile Radio Communications.

A.C.O. Azubogu, C.O. Ohaneme, S.U. Ufoaroh and S.U. Nnebe Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria [email protected], [email protected], [email protected], [email protected]

ABSTRACT Propagation channel measurement can be conducted to study the space-time channel characteristics. However, channel measurement is a time consuming work and often limited by the resolution and complexity of equipment. In contrast, as a deterministic prediction tool, ray tracing simulation methods are computer based and can provide more channel information than measurements. In this paper, a 3-D Radiowave Propagation Simulator (RPS student) developed by Radioplan is used as a wireless channel simulation tool to characterize a wideband propagation channel in terms of received signal strength from which we computed the path loss exponent of the environment, the power delay profile which enable us to determine the mean delay and root mean square delay and power angle profile of microcellular radio propagation at 800MHz band for Awka suburban environment in Nigeria. The predicted values of the received signal strength were compared with measurement conducted at that site and found to be in agreement. The spatial and time dispersive results from the simulation were compared with values from similar environment in literatures..

Keywords- Ray tracing, power delay profile, pathloss exponent

1. INTRODUCTION

Propagation of radio waves through wireless channels has been an They are also very useful for performing interference studies as interesting phenomenon in telecommunication, especially in the deployment proceeds. These models can be broadly mobile communication system. The emergence of wireless system categorized into three categories; empirical, stochastic and and its attendant constraints in terms of the nature of propagation deterministic models. Empirical models are those based on channels has provided much needed research in determining the observations and measurements alone. These models are mainly mode of signal propagation through various channels. The used to predict the path loss, but models that predict multipath characteristics of wireless channels in different propagation parameters have also been proposed [1]. Stochastic models, on the environments determine the nature and level of the signal powers other hand, model the environment as a series of random transmitted in a particular environment. This paper, therefore, variables. These models are the least accurate but require the least explores the existence of various channel impairments at different information about the environment and use much less processing propagation environment, to develop a channel characterization power to generate predictions. The deterministic models make use technique that will help in simulating mobile radio system for of the laws governing electromagnetic wave propagation to optimum propagation results. Hence, the deployment of ray determine the received signal power at a particular location [2]. tracing technique in this work leaves no stone unturned in a There are quite a number of literatures on deterministic ray tracing successful characterization of wireless radio channel through simulation for characterizing propagation channels for both simulation in order to alleviate the problems that occur within a indoors and outdoors. Review of these papers really provided communication medium. evidence for the reliability of this modern method of propagation modeling. Some of these papers are reviewed in this section. [ For the channel characteristics to be determined, signal strength measurement is carried out to ascertain the levels of transmitted 3] investigated the prediction accuracy of an advanced signals in a particular propagation environment. The propagation deterministic propagation model in terms of channel environment can be either of indoor or outdoor. To determine the depolarization and frequency selectivity for indoor wireless received signal strength in any environment, a ray tracer is propagation. In addition to specular reflection and diffraction, the deployed to track the transmitted signal from base stations which developed ray tracing tool considers penetration through dielectric are meant to be received by various mobile stations. Also in this blocks and or diffuse scattering mechanisms. The sensitivity and work, propagation models are developed to study the propagation prediction accuracy analysis is based on two measurement of radio signals through such environments using simulation. The campaigns carried out in a warehouse and an office building. It is models are simulated using ray tracer to characterize the wideband shown that the implementation of diffuse scattering into RT channels in mobile radio system. Propagation models are used significantly increases the accuracy of the cross-polar extensively in network planning, particularly for conducting discrimination prediction, whereas the delay-spread prediction is feasibility studies and during initial deployment. marginally improved.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Besides, [4] proposed ray tracing deterministic method for mobile 2. DETERMINISTIC PROPAGATION MODELLING communication channel modelling due to its ability to analyze The deterministic models make use of the laws governing frequency selective, time invariant as well as spatial behaviour of electromagnetic wave propagation to determine the received signal the channel. They presented three application methods using a ray power at a particular location. Deterministic models often require tracing model (developed at Institute for High-frequency a complete 3-D map of the propagation environment. Depending technology and Electronics, Karlsruhe University, Germany) to on the modelling approach, they can provide both narrowband and characterize the channel parameters. First, the characteristic wideband analyses including path loss exponent, delay-spread and channel parameters for a high speed train communication system angular-spread of the propagating waves. A good application of were presented. The fast moving trains cause a high Doppler deterministic modelling is the ray tracing [1]. spread in the channel which is a special challenge for system designers. Second, the same ray tracing tool was used to calculate 2.1 Ray Tracing the propagation channel in urban outdoor environment. Third, a Ray tracing is based on a detailed simulation of the actual physical car-to-car communication system was modeled. The model was wave propagation process. In order to produce a deterministic verified by comparing measurements with simulation results in a description of the wave propagation, suitable formulation of the realistic urban scenario and in statistically generated high way physical propagation phenomena are applied to a deterministically scenarios. Algorithms to improve the efficiency and accuracy of described scenario. The modelling of propagation phenomena is ray tracing deterministic methods have also been proposed in usually based on geometric optics (GO) and uniform theory of literatures. diffraction (UTD) [7]. Ray tracing describes each multipath component found in the A novel quasi three-dimensional (3-D) ray tracing (RT) algorithm scenario by a ray. A distinction is made between different wave was proposed by [5] which takes into account the advantages of propagation phenomena by deriving several transmission paths. A both the image theory (IT) and the shooting-and- bouncing ray detailed description of the propagation environment (size, position (SBR) method. It is based on creating a new virtual source tree in and composition of materials) is essential for an accurate channel which the relationship between neighbour nodes is a left-son-and- prediction. Several commercial software have been developed for right-brother one. Their theoretical results of the signal path loss propagation channel prediction such as RPS developed by along the streets were compared with measurements which have Radioplan, Ericson EDX, X-Siradif from SIRADEL, QEDesign, been reported for city streets in Tokyo and Ottawa City for various PlaNet, RAPSOR, winprop, ray tracer 2000 etc. values of the propagation parameters. The good agreement with these measurements indicates that their prediction model works 3. WIDEBAND CHANNEL PARAMETERS AT 800MHZ well for such microcellular communication applications. USING RPS In this work, we characterized a wideband channel via a ray Also, [6] proposed ray tracing technique for characterizing an tracing simulation with the goal of extracting parameters of indoor wireless system. This system is used in large variety of interest. A 3-D software known as Radio Propagation Simulator office, factory and residential building. Thus adequate guidelines (RPS) is used for this prediction. The parameters of interest are for radio port placement are needed to ensure satisfactory chosen because they have important implications when designing performance at the lowest cost. They described an efficient 3-D a communication system. Average received power is often used as ray tracing technique algorithms which account for all rays a determination of link quality, while delay spread determines the (transmitted as well as reflected) rays reaching the receiver maximum data rate possible without compensating for the effects location after an arbitrary number of reflections. They also of Inter-Symbol Interference (ISI). Angular spread is an important included the effect of angle of incidence, the material dielectric parameter for designing receivers that employ spatial diversity. constant and the antenna patterns. This technique was applied to The spatial cross-correlation function determines how far diversity the AT&T Bell Laboratories facility at Crawford Hill. The major antennas must be separated before the fading of their received goal of this paper is to present and test a 3-D ray tracing voltage envelopes becomes uncorrelated [8]. simulation approach to characterize the wideband propagation channel in terms of received signal strength and the power delay 3.1 Ray Tracing Tool – RPS profile of micro-cellular radio propagation at 800MHz band. RPS, the Radiowave Propagation Simulator, is a radio coverage/performance planning software for a variety of radio The rest of this paper is organized as follows: Overview of systems. This software is developed by Radioplan GmbH, deterministic model is presented in section two. Section three Dresden, Germany. Before a simulation can be performed, an presents the methodology behind ray tracing simulation for environment has to be generated using the integrated environment prediction of wideband parameters using RPS. Section four editor or be imported from CAD tools such as AutoCAD [9]. present the result analysis. Finally, we present the conclusion and Next, the radio network is configured by placing transmitters at future work on the topic in section five. selected base station positions, whereas receivers are usually set in a matrix or along a line at places where channel data shall be obtained. Then, a network simulation can be performed.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

3.2 Description of the Tested Environment 3.3 Network Configuration The site of interest in this work is the suburban region of Awka, A network configuration in RPS is made up of the arrangement of specifically the Anambra State secretariat complex and its transmitters and receivers together with their associated antennas environs. A case study is the Visafone base station located at the and parameters such as carrier frequency, transmitter antenna nearby police station shown in Figure 1 while the equivalent height, receiver antenna height, transmitting power. A receiver database is created as shown in Figure 2. Measurements of the represents a location where the channel impulse response is received signal strength of this base station have already been observed. By arranging receivers along a line or in a matrix grid, taken [2] which will validate the result of the simulation. the spatial distribution of the field strength can be efficiently determined. A receiver is characterized by its position and its height. All receivers in a network configuration have an associated common antenna with certain orientation and polarization. Table 1 gives the summary of the network parameters used in carrying out our simulations. These parameters are used in the RPS network setup.

Table 1: RPS Simulation Parameters Type of Tx antenna Sector (tilted 300) Tx antenna height 38m Transmit power 41.4dBm Polarization Linear vertical Carrier Frequency 878MHz Rx antenna Type Isotropic Rx antenna height 1.5m Noise Floor -160dBm Ray Tracing Algorithm 3-D

Figure 1: Site of interest (Testbed environment) 4. RESULT ANALYSIS Figure 1 shows the picture of the simulated environment. The environment composed of few buildings and more trees and The results derived from RPS simulation can be presented in vegetations. The average building height is 15 meters. The length surface plots, graphs and tables and be analyzed for a given of the area where measurements were taken is about 2km. position (point analysis) and along a specific path (path analysis).

4.1 Point Analysis This provides the impulse response at a receiver point in 2D view as well as the Direction of Arrival (DoA). The channel impulse responses are the fundamental results in time domain generated by the ray tracing algorithms [9]. Most of all channel parameters can be derived from them. A channel impulse response contains impulses with a given delay, and a complex magnitude. Channel impulse response at a receiver point derived from simulation is shown in Figure 3.

Figure 2: Environment Database

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 3: Channel Impulse Response at a Receiver Point Fig 4b Direction of Arrival theta-plane

4.2 Path Analysis 4.1.2 Direction of Arrival In many cases, a result analysis along a path is very useful to understand the Line of Sight (LoS)/ Non Line of Sight (NLoS) Every ray hits a receiver with a specific incidence angle, called behaviour in a certain environment [9]. RPS has built-in path direction of arrival (DoA) or angle of arrival (AoA) as shown in analysis functions to investigate the received power and delay Figures 4a and 4b. On the other hand, the outgoing angles at the spread along a given path in 2D view as shown in Figures 5a and transmitter can also be of interest. RPS stores directions of arrival 5b respectively. at both transmitter and receiver objects in the impulse responses. The spatial resolution in the related charts is set to one degree. All magnitudes of rays that cannot be separated in space are added incoherently. In analogy to the channel impulse response chart, all empty points are replaced by the noise floor value.

Fig 5a: RSSI along the specified path

Fig 4a: Direction of Arrival phi-plane

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Matlab is used to analyze both Table 1 using equation 1 to obtain path loss exponents of 2.68 and 2.62 for simulation and measurement respectively.

4.4 Comparison with Real Measurement To validate the result of simulation, comparison between simulation and real measurements for the received signal strength is given in Figure 5.

Received Power along path -50 Measurement -55 RPS

-60

-65

Fig 5b: Delay Spread along the path -70 -75

4.3 Determination of Path Loss Exponent -80 To truly characterize propagation path loss for a specific cell site, values should be established for the path loss at a close-in [dBm] Received Power -85 reference distance, Lp(d0) and path loss exponent, n. -90

Using a reference distance of 100m, Lp(d0) is 101dB. The path loss -95 exponent n, which characterizes the propagation environment in -100 our site of interest, is obtained by method of Linear Regression 0 200 400 600 800 1000 1200 1400 1600 1800 (LR) analysis. The path loss exponent n is given as [2]: Relative Distance [m] K Figure 6: Comparison between Simulation results and real LP di  LP d0  Measurements. n i1 (1) K  di  From Figure 6, it is obvious that there is a good agreement 10log  between the simulation and real measurement. i1  d0  Table 2 shows the data created from the received signal strength from the simulation as well as measurement at interval of 100m. 4.5 Determination of the Delay Spread 3-D for the power delay profile from RPS is shown Figure 7 Table 2: Received Signal Strength (RSS) [dBm] Simulation Measurements Dist. Rx Average Rx. Average From Tx power Path loss Power Path loss [m] [dBm] [dB] [dBm] [dB] 100 -60 101 -62 103 200 -60 101 -65 106 300 -77 118 -52 93 400 -83 124 -53 94 500 -82 123 -81 122 600 -74 115 -83 124 700 -88 129 -80 121 800 -89 130 -94 135 900 -94 135 -91 132 100 -89 130 -93 134 1100 -85 126 -94 135 1200 -92 133 -98 139 1300 -91 132 -98 139 1400 -89 130 -99 140 Figure 7: Power delay profile along the path 1500 -87 128 -97 138 1600 -86 127 -96 137 1700 -85 126 -100 141

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

From the data created in the result analysis, time dispersive characteristics of the propagation channel (i.e. mean delay and RMS delay spread) can be computed using equations 2 and 3 respectively. Mean delay,  0 and rms delay spread  RMS is given as [10]: 1 n  0   pi i (2) pT i1 n where PT Pi i1 n 1 2 2   P  (3) RMS  i  i  0 PT i1 The predicted values for mean delay and rms delay spread are given in Table 3. The time characteristic of the environment is in the range with results in other literatures for suburban Figure 8: Power angle profile at the receivers environments [10,11,12,13].

Table 3 Delay Spread 5. CONCLUSION

Parameter Values in s  In this work, we have presented an efficient ray tracing computer simulation propagation model to study a wideband channel 1.1733 characteristics of wireless propagation at 800MHz band for Awka Mean delay ( 0 ) suburban environment. We have shown that, with acceptable computation time, it is possible to compute important wideband 0.5539 channel parameters that are necessary to characterize this RMS delay spread ( )  RMS particular environment.

Path loss exponent of 2.68 was obtained from simulation. This is 4.6 Determination of Angle Spread an important parameter for pathloss model. The power angle profile or power azimuth spectrum at the receivers is shown in Figure 8. The mean delay and RMS delay spread are 1.1733 and From the data created in the result analysis, spatial characteristics  sec of the propagation channel (i.e. mean arrival angle and RMS angle 0.5539  sec respectively. Such representation can give spread) can be computed using equations 4 and 5 respectively. precious information to calibrate digital transmission systems, Mean arrival angle,  and rms angle spread  is given as particularly the number of fingers to take into account in a Rake 0  receiver including the maximum symbol rate which can be [14]: transmitted over the multipath radio channel with moderate distortion. Finally, we computed the mean arrival angle and the N N 0 0 P  P  RMS angle spread to be 163 and 58.65 (approximately 1rad)  n n  n n respectively. Angle spread is an important parameter for designing n1 n1 0  N  (4) receivers that employ spatial diversity. PT Pn n1 5.1 Future Work The accuracy and reliability of any ray tracing computer  N   P  2  simulation model depend to a large extent on the accuracy of the  n n geographical database representing the actual propagation    n1  2  (5)   P 0  environment. In order to achieve an accurate database creation,  T  photogrammetric software should be developed or a reliable   existing one be used. These softwares are capable of converting The predicted values for the mean arrival angle and rms angle the photograph of the environment into a digital format, hence a spread are 1630 and 58.650 respectively. Angular spread of 58.650 true replica of the actual environment. Our future work should also (approximately 1rad) indicates that arriving power is not biased in focus on using this ray tracing tool (RPS) on different a particular direction [15]. environments in order to validate its result.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

REFERENCES Authors’ Profile

[1] Erceg V., Hari V. S., “Channel Models for Fixed Wireless Applications,” Tech. Rep., IEEE 802.16 Broadband Wireless Augustine C. Okwudili Azubogu holds Access Working Group, 2001. PhD in Communications Engineering [2] Azubogu A.C.O., Onoh G.N., Idigo V.E., Ohaneme C.O., from department of Electronic and “Empirical-Statistical Propagation Path loss Model for Computer Engineering, Nnamdi Azikiwe Suburban Environment of Nigeria at 800MHz Band”, IUP University, Awka, Nigeria. He is currently Journal of Science and Technology(IJST), Hyderabad, a lecturer in the Department of Electronic India, Vol. 7, No. 2. 2011. and Computer Engineering Nnamdi [3] Mani F., Oestges C., Quitin F., “Accuracy of depolarization Azikiwe University PMB 5025 Awka, Nigeria. His research and delay spread predictions using advanced ray-based interest is on wireless communication. E-mail: modeling in indoor scenarios”, EURSIP Journal on Wireless [email protected] Communication and Networking, 2011. [4] Knorzer S., Fugen T., Wiesbeck W., “Ray Tracing for Mobile Communications”, Proceedings of WFMN07, Chemnitz, Germany, 2007, pp 10-15. [5] Liu Z.-Y., Guo L.-X., “A Quasi Three-Dimensional Ray Cletus Ogbonna Ohaneme holds PhD Tracing Method Based on the Virtual Source Tree in Urban in Telecommunication Engineering from Microcellular Environments”, Progress In Electromagnetics the Department of Electrical/Electronic Research, Vol. 118, 2011, pp 397-414. Engineering, Enugu State University of [6] Valenzuela R.A., “Ray tracing Prediction of Indoor Radio Science and Technology (ESUT), Enugu, Propagation”, PIMRC, AT&T Bell Laboratories, Crawford Nigeria. Presently, he is a lecturer in the Hill Laboratory, Holmdel, NJ 0733, 1994, pp 140-142. Department of Electronic and Computer [7] Smulders P., Jevrosimovic M., Herben M., Savov S., Martijn Engineering, Nnamdi Azikiwe E., “State of the art channel models”, Broadband Radio at University PMB 5025 Awka, Nigeria. Hand 2002. His research interest is on Mobile Wireless System and Spectrum [8] R.G. Vaughn and N.L. Scott, “Closely Spaced Monopoles for Management. E-mail: [email protected] Mobile Communications”, Radio Science, Vol. 28, No. 6, 1993, pp. 1259-1266. [9] Deißner J., Hübner J., Hunold D., Voigt J., “RPS Version 5.4 Stephen Uchenna Ufoaroh holds M.Eng. User Guide”, Actix GmbH, Altmarkt 10 D-01067 Dresden, in Communications Engineering from Germany 2008. Nnamdi Azikiwe University Awka, Nigeria [10] Saunders S. R., Alejandro A., “Antennas and Propagation for and currently a doctoral candidate in wireless Communication Systems”, ed. 2, John Wiley & Computer and Control Engineering in the Sons Ltd, The Atrium, Southern Gate, Chichester, West Department of Electronic and Computer Sussex, PO19 8SQ, England 2007. Engineering, Nnamdi Azikiwe University [11] Sousa E.S., Jovanovik V.M., Daigneault C., “Delay Spread Awka, Nigeria. He is also a lecturer in the Measurements for the Digital Cellular Channel in Department of Electronic and Computer Toronto”, IEEE Transactions on Vehicular Technology, Engineering, Nnamdi Azikiwe University Vol. 43, No. 4, 1994 pp 837-842. Awka, Nigeria. His research interest is on Software and Control [12] Rappaport T.S., “Wireless communications: principles and Engineering. E-mail: [email protected] practice”, Prentice-hall of India, New Delhi 2008. [13] Katz R. H., “Radio Propagation”, Lecture Note, University of California, Berkeley, CA 94720-1776, 1996, pp. 10. Scholastica Ukamaka Nnebe holds [14] Gross F.B., “Smart Antenna for Wireless communications”, M.Eng. in Communications Engineering McGraw-Hill Inc, 2005, pp 145 - 148. from Nnamdi Azikiwe University Awka [15] Patwari N., “Measured and Modeled Time and Angle Nigeria. She is currently a doctoral Dispersion Characteristics of the 1.8 GHz Peer-to-Peer candidate in Communication Engineering Radio Channel”, M.Sc Thesis, Faculty of the Virginia in the Department of Electronic and Polytechnic Institute and State University, Blacksburg, Computer Engineering, Nnamdi Azikiwe Virginia 1999. University Awka, Nigeria. Presently, she is a lecturer in the Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria. Her research interest is on Wireless Sensor Networks and System Engineering. E-mail: [email protected]

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Engendering Quality Information Systems Development through Software Architectural Design

1GAMBO Ishaya, 2ACHIMUGU Philip, 3IKONO Rhoda, 4IROJU Olaronke, 5SORIYAN Abimbola 13&5Department of Computer Science and Engineering, Obafemi Awolowo University, Ile-Ife, Osun State 2 Department of Computer Science, Federal University of Agriculture, Abeokuta 3 Department of Computer Science, Adeyemi College of Education, Ondo State, Nigeria. [email protected], [email protected], [email protected]

ABSTRACT Quality issues have been a fundamental focal target in the development of most information systems. This is because most stakeholders involve in the usability of Information Systems(IS) are particular about quality in terms of user interface, functionality, ease of use and satisfaction. For a robust interactive society, quality information systems (QIS) should provide all the quality-relevant information during the whole life cycle of a product to all stakeholders. In this paper we are of the opinion that attention needs to be paid to architectural design which can be used to predict quality issues in the process of IS development thereby improving the quality of interaction within the IS. The paper present a framework to show how architectural design as a software engineering strategy can be used to predict quality of systems and also use as a communication vehicle among stakeholders for a better interactive society.

Keywords- Information systems, architecture, software engineering, quality, life cycle

1. INTRODUCTION

IS quality can be synonymous with “error free” system. In Consequently, software engineering is an important practice in today’s ISs practice and organization, the software systems used developing and maintaining information systems. Within the offers a great level of supportability in the daily discharge of engineering practices, the architectural design is very crucial in duties and on the overall activities within an organizational the attainment of quality goals. Quality issues have been a setting or structure. Software systems are majorly indispensable fundamental focal target in the development of most information elements since they support the quality of services rendered in systems. This is because most stakeholders involve in the order to ease workload and make the organization more usability of information systems are particular about quality in effective. Even in some of the hardware components used by the terms of functionality, ease of use and satisfaction. In this paper people within an information system setting for interaction, we are of the opinion that attention needs to be paid to these software systems are hidden in the hardware architectural design which can be used to predict quality issues products/devises they use. The software controls vital functions in the process of information systems development for a better in the management of most information systems for maximum interactive society. delivery of quality services and for the purpose of improving the effectiveness and efficiency of the system. Most often these software systems are developed with a particular purpose. For example, it provides specific In fact, the economic importance of software element in an functionality which supports or replaces a person to do a certain information system is incontestable as well as the dependence of task in a specific context. Functionality and quality attributes are society on software for better interactivity. Many services that orthogonal. Functionality can be achieved through any number we take for granted would have been impossible to create of structures because it is largely independent of structures, without the use of this software. With this in mind, we believe which is being constrained by the quality attributes. Achieving that the functionality and parallel versions of products realized these quality attributes must be considered throughout the by software have increased greatly along with the rising development process of software systems for an information significance of software quality. Therefore, an increasing need system. In the context of this paper, software quality is for faster, cheaper and even more versatile software products considered as the degree to which software possesses a desired sets a real challenge for the software industry. This is why the combination of attributes (e.g. performance, security, portability, software industry is constantly looking for ways to improve the usability, and efficiency, etc.). We are of the opinion that cost-effectiveness of software development and the quality of quality attributes of information systems can be highly software products in order to support and control the constrained by a system’s software architecture or information management of information systems. system architecture. Thus, this paper discusses the time an information system’s architecture is specified; when the system will have the desired qualities and when it will follow any specific architecture for the quality to be determined at all times.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

2. QUALITY ISSUES IN INFORMATION SYSTEM On a general note, emphasizing processes in information DEVELOPMENT systems development has contributed significantly to the understanding of how such systems should be developed. We are of the believe that the process (development methodology) is Quality in the domain of IS is a relatively new research what determines the quality product of the system. During the direction. Quality issues are paramount to all stakeholders process, interaction can be enhanced, while the product can be involve with the use of information system in any given and used as means to improve the interaction. Basically, before established organization. This has been the on-going research Information Systems are developed, they must have undergone a efforts for many in the software engineering community. process called Systems Development Life Cycle (SDLC) using Basically [1] gave two major camps when discussing the appropriate methodology. meaning and definition of (software) quality. According to their work, “quality means conformance to specification and also The SDLC consist of phases varying from author to author. meeting customers’ needs”. Putting it in a more comprehensive However, an information systems project can only be successful and technical way, we have lots of quality strategies proposed with intense interaction amongst project manager, systems by [2], [3], [4], [5], [6], [7], and models by [8], [9], [10], [11], analyst, system designers and the end users. Viewed from these [12], [13], [14] etc. All the quality models seem to be a good different stakeholders perspective, the SDLC need to lay more reference tool for defining the quality of an information system. emphasis on the architectural design at all levels to ensure In fact, in the computerized society we are in, these qualities are quality satisfaction. The focus on web based IS development increasingly important. In a way also, the quality attributes are have been the present and current change in IS development. the realizations of a product or service quality. That is, quality These includes both portal and intranet development. Some of attributes are the specific information system characteristics that the examples include Web Information System Development combine to produce system quality. Method (WISDM) [20], Object Oriented Hypermedia Design Chung provides an extensive listing and description of quality Method (OOHD) [21], [22] the Intranet Development attributes in Non-Functional Requirements in Software Methodology (IDM) [23], and other approaches that are based Engineering [15]. Quality attributes arise in the business case, on rapid application development, such as agile system market analysis, and requirements, and can refer to the product development [24] and extreme programming by Beck [25], have being developed or to the development of that product [16]. become prominent. In this paper we admit that new information Such quality attributes tend to be global, referring to the entire system development typically starts with a temporary product or the development of that product. For example, a organizational structure called project team. Typically, a project business might have goals of producing products that are easy team consists of a project manager, system analyst, for customers to use and of producing new products quickly. programmers, etc. In the novelty of project team, it will be nice 2.1 Information Systems and Information Systems and adequate to have the architect whose role is strictly to Development design the architecture, which serves as the blue print for the entire system. Different definitions on information system exist today in various context viewed from different perspective. In this paper 2.2 Software Architectural Design and Information Systems the academic discipline-point of view on information system Architecture takes into cognizance the view from the real world perspectives The building block of any kind of software system is its and practices in organizations since interaction largely takes architecture. The architecture determines the degree of success place at the organizational level. According to Ciborra [17], of a system. Software architecture in software engineering (SE) “information system is a system which includes the technology practices was introduced to help manage complexity and risk and the people, and the whole context where the system is”. In during development and maintenance of software systems. It the context of development project, information systems are guides and documents the structure of the system as well as the often understood as covering only the technology, such as role each system’s components play within the structure. It equipment, methods, and practices, which can cause a technical embodies the earliest software design decisions. These decisions bias in implementation because the focus on the human enable or preclude the achievement of a desired system quality, environment and people is inadequate [18]. such as reliability, modifiability, security, performance and usability etc. Walsham [19] puts it nicely by saying “IS are drawn on to provide meaning, to exercise power, and to legitimize actions” The architecture also forms the basis for the development Putting all these together, it is obvious then to say the focus is approach and acts as the blueprint that guides the teams building strictly on the people and the technology the people use. In this the system. Software architecture (SA) is part of the information case adapting information systems should be in a dynamic system architecture (ISA). The ISA is a part of a vaster field of environment, which is the major difference in the shift from architectures and models relevant for the organization. traditional engineering focus on product to an emphasis on Considering the architectural scope and level, we can distinguish process. This shift so far has brought about the development enterprise architecture, information systems architecture and strategies and methodology employed on information systems software architecture. practices.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

To appreciate the fact that technology support business needs This includes the information architecture or data architecture and that IT investment deliver value, it is fundamental to define that represents the main data types that support business [29], the role software architecture plays with the information systems [30]; the application Architecture that defines application architecture. According to Carnegie [26], software architecture needed for data management and business support; and the main study area is on how programs or application components Technological Architecture that represents the main are internally built. Within an information system architecture, technologies used in application implementation and the software architecture addresses the representation of the IS infrastructures that provide an environment for IS deployment – components structure, it relationships, principles and directives as network, communication, distributed computation, etc. [29], [27]. [28] added by saying such representation is with the main [31]. All of these can actually support the need to good propose of supporting business. More so, the information system architectural design for quality satisfaction. architecture according to [29] distinguishes three aspects that define three sub architecture.

3. PROPOSED ARCHITECTURAL DESIGN FRAMEWORK FOR IS DEVELOPMENT

Figure 1: Proposed Information Systems Development Model

Today, modern approaches to software architecture have taken a The use of multiple views allows to address separately the multi-view approach. However, some literatures like [32], [33] concerns of the various ‘stakeholder’ of the architecture, and to and [34] agree at a point that for a software architecture handle separately the functional and non-functional description, different views are necessary for describing the requirements. If we consider the analogy of the architecture of a different aspects of a software system. Among these views are building, for example, various stakeholders (such as the the description of the static organization of the software system construction engineer, the plumber, and the electrician) all have and the development environment as well as the description of an interest in how the building is to be constructed. Although the hardware architecture. A view is the representation of one or they are interested in different components and different more structures present in the system. relationships, each of their views is valid.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Each view represents a structure that maps to one of the In practice, architecture design is typically seen either as a architecture of the construction goals of the building, and these separate task, which can be executed before the analysis/design multiple views are necessary to represent the architecture of the phase, or as an otherwise separate task which needs different building fully. Similarly, software architecture has varieties of kind of knowledge, skills, and tools than analysis and design stakeholders, including developers, maintainers, testers, normally does. Major part of the research on software integrators, system administrators, project managers, analysts, architecture design seems to fall on the following categories: certification authorities, and end users (all these are also included in the SDLC). 1. Architecture description languages and component/connector models. Each of these stakeholders has a vested interest in different 2. Architectural styles and patterns. system properties and goals that are represented by different 3. Modifications to modeling languages to support structural views of the system. The views provide the basis for architecture design. reasoning about the appropriateness and quality of the architecture for achieving system quality goals. The views also However, information systems development (ISD) is a change address one specific set of concern using several concurrent process, which aims at changing the object system according to views. However, if the architecture is not right, the system will some objectives. The object system consists of the phenomena not meet its requirements. Software architecture has been perceived by the development group and it identifies the target identified as an increasingly important part of information of change in the ISD process. A development group carries out system development. The architecture helps the developer to the process of change and at least partly shares a common define the internal structure of the system. The requirements language and a concept structure to enable communication and towards information system usually go beyond the correct common understanding. They make perceptions on the object functionality, the presence of certain quality demands are also system, act to change it, and describe the system using the very essential for the systems' acceptance by the stakeholders description languages and conventions offered by the ISD and users. So, quality control and management must be carried approach and method. The final aim of the development group is out through the whole development process to ensure the to implement the change in the user organization or community implementation of required quality characteristics. Quality according to the stated or specified objectives. This change attributes of information systems can be highly constrained by usually includes composition and installation of formal artefacts architectural design. Thus, it is good to pay attention to (such as software and hardware) and changes in the rules and architectural design and specification so as to have the desired practices in the user organization or community as depicted in qualities [35]. figure 1. The whole process takes place in a set of environments, including organizational, technical, and social environments. The key to overcome the complexity and change of a complex product, such as an enterprise or an information system, is Consequently, we are of the opinion that an explicit picture of architecture. If the product is so complex that its author cannot the software architecture needs to be defined. In our thinking, remember all details, he/she has to write down its architecture. the software architecture should be clearly defined at the In the last years, much attention has been paid to software problem and solution context within the proposed model in architecture. It seems that people need strong analogues and figure 1. The architecture should be part of the technical metaphors from the physical world to understand collectively environment which can have effect on the change process, and the structure and the organization of the software domain. should also form the solid part of the object system that must be Despite this unclear situation and an undefined relationship perceived, described and changed. Therefore, when designing an between information systems development and software architecture within the Information System Development architecture design, there are many practical reasons that make framework, we are structuring the solution, which has great software architecture design a central issue in information influence on the quality. The essential parts of the architecture systems development. For example, distribution, platform are the components, their interfaces and connections, modules, heterogeneity, and security are major issues in mobile and web- and recurring software patterns and policies that are determined based information systems necessitating rigorous and well- by the architecture. The relationship that exists among the designed software architecture. In addition, fast time-to-market interfaces and connections, modules etc. are also essential. This and sufficient quality are conflicting but necessary requirements is depicted in figure 2. for commercially successful software development. Well designed software architecture plays also a major role in In figure 2, the IS object system framework was divided into the resolving this conflict by enabling reusability and rapid change. IS analysis and IS architectural design models. The problem context is embodied in the IS analysis model that imposes As a result of this observed importance of software architecture, requirements and application domain concepts to the IS we need to clarify the role of software architecture and software architectural design model, which embodied the solution architecture design in information systems development. In context. The architectural design model describes the software figure 1; we aim to put software architecture design into a larger in the solution context. In this sense, the architecture is not only methodical context and demonstrate how software architecture important in terms of solution and design, but it is also important design and information systems development intersect. because it guides the information systems development process

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

by offering possibilities and setting constraints on how the In a way, architectural design serves as the blueprint for problem context may be perceived when keeping the ISD information system development practices with the view of objectives in mind. Consequently, the architecture forms the predicting and enhancing the quality. Therefore, quality issues in visible part of the solution context when perceiving the problem terms of control, prediction and management in the development context. We are of the opinion that ISD practices needs an process of IS can be greatly engendered through architectural architectural description because it must collectively understand design. The architectural design will ensure the implementation in general terms how the problem is transformed to the solution of required quality characteristics, which mostly are the to enable a common language and a concept structure. One other requirements of stakeholders. possibility of the architectural design is the provision of understanding and communication between different stakeholders participating in the ISD process, which mostly are the primary objectives.

IS Architectural Design Model

Embodies

IS Solution context Imposes requirements on

Offers possibilities and IS Analysis sets constraints on Model

Determines/influence

Embodies Concern IS Quality Assurance

IS Problem Context

Stakeholder(s)

Figure 2: IS Object System Framework

4. FUTURE WORK 5. CONCLUSION

The space between an architectural model and IS quality model When designing the architecture of an information system, so needs to be explored properly; to be able to evaluate up to what many factors are taken into consideration. Some of these factors extent a specific architecture supports the relevant quality include the components, their interfaces and connections, modules, characteristics in a given information system development project. and recurring software patterns and policies that are determined by More so, in considering quality issues in information systems the architecture. The patterns and views (architectural) are good development, more work is needed in the development of a generic reference point to predict the quality when thoroughly analyzed. model for the different perspective of quality requirements from The place of architecture in respect to the problem and solution stakeholders. Since quality is in the eye of the beholder, it then contexts is depicted in figure 2. The object system framework is means more work is needed to show the relationship that could divided here into the IS analysis and IS Architectural design exist among the different perspectives. models. The problem context is represented in the analysis model that imposes requirements and application domain concepts to the IS architectural design model representing the solution context.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Consequently, software architectural design during information 15. Chung, L. Nixon, B. Yu E. and Mylopoulos, J. Non- systems development will have great influence on the quality of functional requirements in software engineering. Kluwer the system. The architecture gives the ISD practice an Academic, Boston, 2000. understanding and common knowledge about the system’s 16. Chastek, Gary & Donohoe, Patrick. Product Line technical environment and offers possibilities and sets constraints Analysis for Practitioners (CMU/SEI-2003-TR-008). on the perception of the problem and on the construction of the Pittsburgh, PA: Software Engineering Institute, Carnegie solution. When this is done, quality issues will be attained or Mellon University, 2003. achieved. Architectural practices during ISD also help in defining 17. Ciborra, C. (2004): Encountering information systems as recurring software patterns that may be used as economical a phenomenon. In Avgerou, Ciborra, Land (eds.). The modeling constructs in analysis and design of information systems. Social Study of Information and Communication As such, it can be concluded that quality control and management Technology. Oxford University Press Norfolk 2004, pp must be carried out through the whole development process of 16 – 37. information systems to ensure the implementation of required 18. Jacucci, E., Shaw, V., Braa, J. (2006). Standardization of quality characteristics. Health Information Systems in South Africa: The Challenge of Local Sustainability. Information Technology for Development. Vol 12/3. REFERENCES 19. Walsham, G. (2000) Globalization and IT: Agenda for Research. In Baskerville, R., Stage, J. and DeGross, J. 1. Hoyer, R., and Hoyer, B. (2001). "What is quality?", (eds.) Organization and Social perspectives on Quality Progress, no. 7, pp. 52-62. information technology: IFIP International Working 2. Crosby, P. (1979). Quality is free : the art of making Conference on the Social and Organizational quality certain, New York : McGraw-Hill. Perspective on Research and Practice in Information 3. Deming, W. (1988). Out of the crisis : quality, Technology (TCB WG8.2), 9 – 11 June, Aalborg, productivity and competitive position, Cambridge Univ. Denmark. (Boston: Kluwer Academic Publishers 2000), Press. pp 195 - 210 4. Feigenbaum, A. (1983). Total quality control, McGraw- 20. Vidgen, R., Avison, D., Wood, R., and Wood-Harper, A. Hill. T., Developing Web Information Systems, Butterworth- 5. Ishikawa, K. (1985). What is total quality control? : the Heinemann, London, 2002. Japanese way, Prentice-Hall. 21. Schwabe, D., and Rossi, G., “The Object-oriented 6. Juran, J. (1988). Juran's Quality Control Handbook, Hypermedia Design Model,” Communications of the McGraw-Hill. ACM, (38:8), 1995, pp. 45-46. 7. Shewhart, W. (1931). Economic control of quality of 22. Howcroft, D & Carroll, J. (2000). A Proposed manufactured product, Van Nostrand. Methodology for Web Development. Proceedings of the 8. McCall, J. (1977). Richards, P. K., and Walters, G. F., 23. Lee, S., “IDM: A Methodology for Intranet Design,” "Factors in Software Quality", Nat'l Tech.Information Proceedings of the International Conference on Service, no. Vol. 1, 2 and 3. Information Systems (ICIS), Helsinki, Finland, 1998, p. 9. Boehm, B., Brown, J., Kaspar, H., Lipow, M., McLeod, 51-67. G., and Merritt, M. (1978). Characteristics of Software 24. Cockburn, A., Agile Software Development, Pearson- Quality, North Holland. Longman, UK, 2002. 10. Grady, R. (1992). Practical software metrics for project management and process improvement, Prentice Hall. 25. Highsmith, J. A, Agile Software Development 11. Dromey, R. (1996). "Concerning the Chimera (software Ecosystems, Addison-Wesley, Boston, 2002. quality)", IEEE Software, no. 1, pp.33-43. 26. Beck, K. (2000): Extreme Programming Explained. 12. ISO (1997). International Organisation for ISBN: 0-201-61641-6, Addison-Wesley. Standardisation, "ISO 9000-2:1997, Quality 27. Vasconcelos, A., P. Sousa, and J. Tribolet, Information management and quality assurance standards — Part 2: System Architecture Evaluation: From Software to Generic guidelines for the application of ISO 9001, ISO Enterprise Level Approaches, 12th European Conference 9002 and ISO 9003” On Information Technology Evaluation (ECITE 2005), 13. ISO (2001). International Organisation for Turku, Finland, September 2005. Standardisation, "ISO 9126-1:2001, Software 28. Garlan, D., Allen, R., and Ockerbloom, J. (1995). engineering – Product quality, Part 1: Quality model". Architectural mismatch, or, Why it’s hard to build 14. ISO, International Organization for Standardization, systems out of existing parts. In Proceedings of the 17th "ISO 9000-3:1998 -- Quality management and quality International Conference on Software Engineering, assurance standards – Part 3: Guidelines for the Seattle, WA, 1995. Also appears as: Architectural application of ISO 9001_1994 to the development, mismatch: Why reuse is so hard. IEEE Software, 12(6), supply, installation and maintenance of computer pp. 17–26. software (ISO 9000-3:1997)", 1998.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

29. Maes, R., D. Rijsenbrij, O. Truijens, and H. Goedvolk, Redefining Business – IT Alignment Through a Unified Framework, White Paper, May 2000.http://www.cs.vu.nl/~daan/ 30. Spewak, S., and S. Hill, Enterprise Architecture Planning: Developing a Blueprint for Data, Applications and Technology, Wiley-QED, ISBN 0-471-599859, 1992 31. DeBoeve, L. R. (1997). Concept of "Highly Adaptive" Enterprise Architecture. Proceedings of Enterprise Architecture Conference, 1997 32. Open Group, The Open Group Architectural Framework (TOGAF) – Version 8.1, The Open Group, December 2003. 33. Kazman, R., Bass, L., Clements, P. (2003). Software Architecture in Practice. Addison Wesley, second edition. 34. Kruchten, P. (1995). Architectural Blueprints - The .4+1. View Model of Software Architecture. IEEE Software 12 (6), pp. 42-50. 35. Hofmeister, C., Soni, D., and Nord, R. (1995). Software architecture in industrial applications. Proceedings of the 17th international conference on Software engineering. pages 196-207, Seattle, Washington, USA.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Mobility Scenario-based Performance Evaluation of Pre-emptive DSR Protocol for MANET

1V.Ramesh & 2 P.Subbaiah 1Research Scholar, Sathyabama University, Chennai, TN, India. 2Professor in ICE, Sree Sai Ram Engg. College,West Tambaram, Affiliated to Anna University, Chennai, India. [email protected] [email protected]

ABSTRACT Ad hoc wireless networks are characterized by multi-hop wireless connectivity, infrastructure less environment and frequently changing topology. To analyze the performance of routing protocols in MANETs in the real world, a scenario based simulation analysis is required since there is a lack of necessary infrastructure for their deployment. Most of the earlier work done in this field have assumed the Random Waypoint model, which fails to capture the realistic movement of the nodes. In this paper, we describe a set of experiments conducted to analyze the performance of the Preemptive DSR routing protocol in a battlefield scenario. BonnMotion Software(Java based) is used to create and analyses mobility scenarios. Initially an explanation of the experimental metrics and the setup is described, followed by the scenarios used for our simulations. The results give an idea of how the Preemptive DSR protocol behaves in the given scenario and helps identify the metrics for optimal performance of the protocol.

Keywords- MANET, PDSR, battlefield scenario, Packet Delivery Ratio, delay.

1. INTRODUCTION However, node mobility may cause links to be broken frequently, A mobile ad-hoc network (MANET) is a self configuring network of and how to select reliable paths becomes one critical issue for mobile nodes connected by wireless links, the union of which forms routing. Hence, using stable links is crucial for establishing reliable an arbitrary topology. The nodes are free to move randomly and communication paths between mobile nodes. The routing protocol organize themselves arbitrarily; thus, the network’s wireless must react promptly to recover from link and node failures and to topology changes rapidly and unpredictably. Proactive MANET take advantage of new links. For these reasons, existing routing protocols are table driven and will actively determine the layout of protocols designed for fixed networks are unsuitable, and routing in the network. Through a regular exchange of packets meant for MANET is a major issue.[25] network topology between the nodes of the network, a complete picture of the network is maintained at every node. Hence there is 2. DYNAMIC SOURCE ROUTING PROTOCOL minimal delay in determining the route to be taken. The Dynamic Source Routing (DSR) protocol is a simple and Reactive MANET protocols only find a route to the destination node efficient routing protocol designed specifically for use in multi-hop when there is a need to send data. The source node will start by wireless ad hoc networks of mobile nodes[8]. The DSR protocol transmitting route requests throughout the network. The sender will allows source nodes to dynamically finds a route to any destination then wait for the destination node or an intermediate node (that has a node in the ad hoc network. Each data packet sent has in its header route to the destination) to respond with a list of intermediate nodes the complete ordered list of nodes through which the packet must between the source and the destination. This is known as the global pass, and avoiding the need for up-to-date routing information in the flood search, that in turn brings about a significant delay before the intermediate nodes through which the packet is forwarded. DSR packet is transmitted. Since each of the proactive and reactive cache the routing information for future use. DSR protocol contains routing protocols suits well in oppositely different scenarios, there is two major phases, route discovery and route maintenance. good reason to develop hybrid routing protocol that is a mix of both proactive and reactive routing protocols. The hybrid protocol is Route Discovery: It is the process by which a source node S needs to applied to find a balance between the proactive and the reactive send a packet to a destination node D and hence obtains a route to D. protocols[1][2][23]. Route Discovery is used only when source node S needs to send a packet to destination node D, it looks up its route cache to locate an In ad-hoc networks, two nodes communicate with each other in a unexpired route to the destination and if it fails, then it initiates the peer-to-peer fashion. The routes has multiple hops, and hence are route discovery process through broadcasting a Route Request called multi-hop networks. Each node can able to communicate with (RREQ) packet. Each node on receiving a RREQ packet, it the adjacent nodes in its range, and for those which are beyond its rebroadcast the packet to its neighbors if it has not forwarded range, the node takes the help of other intermediate nodes to relay its already. Route Request packet (RREQ) contains [8].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

On receiving the RREQ packet the destination replies the RREQ A. Route Discovery: packet the destination replies to the source with a Route Reply Step 1: When a source node S wants to send a data, it broadcast the (RREP) packet. When an intermediate node detects that the link to RREQ packet to its neighbor nodes. the next-hop node towards the destination is broken, it immediately remove this link from the route cache and returns a route error Step 2: When an intermediate node on the route to the destination message to the source node. The source node again activates a new receives the RREQ packet, it appends its address to the route record route discovery. DSR works for small to medium size MANET when in RREQ and re-broadcast the RREQ. nodes speed is moderate and every node has enough battery power. Its main feature is that every data packet follows the source route Step 3:When the destination node D receives the first RREQ packet, stored in its header. This route gives the address of each node it starts a timer and collects RREQ packets from its neighbors until through which the packet should be forwarded in order to reach its quantum q time expires. final destination. Each node on the path has a routing role and must transmit the packet to the next hop identified in the source route [8]. Step 4:The destination node D finds the two (primary +Backup) best routes from the collected paths (Step 3) within the quantum q time. Route Maintenance: Each node maintains a Route cache in which it stores every source route it has learned. When a node needs to send a Step 5: The destination node D sends RREP packet to the source data packet, it checks first its route cache for a source route to the node S by reversing (RREQ) packets which includes the two routes destination. If no route is found, it attempts to find new route using (Primary +Backup) for further communication. the route discovery mechanism and hereby increases the control overhead and connection setup delay. To overcome these drawbacks B. Route Monitoring: we propose Preemptive Dynamic Source Routing Protocol for Step 1: Each intermediate node on the route starts monitoring the Wireless Ad-Hoc Networks with Backup Route [6][2]. signal strength.

3. PREEMPTIVE DSR Step 2: If signal strength falls below the specified threshold T, it will send a warning message “Path likely to be disconnected”, to the We have proposed this algorithm in the earlier work[8]. source node S. Assumptions: We assume that all nodes wishing to communicate with other nodes C. The Source node S Communicates with destination node D: within the ad hoc network are willing to participate fully in the Step 1: The source node S starts Communicating with destination protocols of the network. Each node participating in the network node D using primary path. should also be willing to forward packets for other nodes in the network. Step 2:On receiving the warning message from the intermediate node, it starts communicating destination node D with the backup We refer to the minimum number of hops necessary for a packet to route also. reach from source to destination. We assume that he diameter of an ad-hoc network will be small(5 to 10 hops), but greater than 1. Step 3: If source node S receives the acknowledgement form the Packets may be lost or corrupted in transmission on the ad-hoc destination node D go to step 4 else step 5. wireless network. A node receiving a corrupted packet can detect the error and discard the packet. Step 4: Preemption, switch over from Primary to Backup route.

Nodes within the ad hoc network may move at any time without Step 5: Initiates Route Discovery Process. notice, and may even move continuously, but we assume that the speed with which nodes move is moderate with respect to the packet D. Prediction Mechanism transmission latency and wireless transmission range of that The main goal of our approach is to avoid sending unnecessary particular network hardware in use. Preemptive DSR can support warning messages. In this work, we consider that anode is in an very rapid rates mobility, but we assume that nodes do not unsafe or preemptive region if the signal it receives from a continuously move with high peed, because it may flood data packet predecessor node is below a threshold signal strength Pt. Once a in ad-hoc wireless networks. The wireless communications link node enters this zone, we make at least three consecutive between each pair of nodes will be bi-directional. But some time the measurements of the signal strength of packets received from the wireless link between two nodes may be uni-directional also. predecessor node, and predict link failure using the Lagrange interpolation. This interpolation has the following general form:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

We store the power strengths of the three signals and their times of 1. Number of intermediate nodes between S(Source) and occurrence. When two consecutive measurements give the same D(Destination) is always n (fixed). signal strength, we store the time of the second occurrence. The 2. Same mobility for each node and hence every link has expected signal strength P of the packets received from the same probability of breakage. predecessor node is computed as follows: 3. Each node moves for a fixed period of time say m, randomly and then remains in rest for a fixed period of time P, which is the pause time. The probability of a node being stable at time T is given by. 4. Stability = Total Pause time up to time T/T. The in stability of a node is, Instability = 1-stability.

Where P0, P1, P2 are the measured power strengths at the F .Mathematical Model measurement times t0, t1, and t2, respectively. The time t is the sum of the time needed for discovering an alternative path Consider a source node S and a destination node D, and a route R (Discovery Period), the last measurement time t2, and the average o from S to D. value of the measurement times t0, t1, and t2. That is:

Probability that Ro will not fail (Fo) = Probability that all of the intermediate links will be stable. n 1 n 1 FSS*So oij oij oj When P is lower than the minimum accepted power (-81 dB) a i 1 i 1 warning message is sent to the predecessor node. This node then j i  1 j  i  1 starts a local repair procedure to find alternative paths to the destinations reached using the link to the node that sent the warning Hence, the probability that Ro will fail = the probability that message. communication link will break. n1 4. MATHEMATICAL ANALYSIS G1  1  F o  F o  1  S oij A. Node Stability : Let gi be the probability that node i is unstable. i1 TotalMotion timeof nodei j i 1 g1 i Totalmotion time of i totalpausetimeof i n1 1 Soi *S oj Pause time is the time duration that the node remains stationary. The i1 node stability is defined as: j i 1 Consider two parallel routes R1 and R2 between S and D. Probability that R will not fail, Sii 1 g 1 n1 B. Link Stability : It is the Probability of the link i-j formed by FS1 1ij nodes i and j is stable. Let S and S respectively denote the stability  i j i1 of nodes i and j. Then the link stability is given by j i 1

n1

Sij S i *S j  1  g i * 1  g j  S*S1i ij i1 C. Path Stability: A path is stable if all the intermediate links are j i 1 stable. Let Fl denote the stability of the path l and Slij. Be the stability of the link i-j along the path l. then we have. n1 Probability that R will not fail, 2 Fl Sl12 *Sl 23 *...... Sln 1 n  S lij i1 j i 1

E. Assumptions : We Assume the following to show that the use of multiple routes provide increased stability.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

n1 In the following sections we describe the experiments carried out to FS analyze the performance of PDSR in a battlefield scenario. It is 2 2ij found that PDSR achieves high packet delivery fraction, low end to i1 j i 1 end delay and normalized routing loads in medium size networks with lower mobility of nodes. n1 S*S2i 2 j Experimental Setup and Metrics i1 The ns-2 simulator was used for the experiments. We now describe j i 1 the traffic pattern, the scenario description and the metrics that were Hence, probability that both R1 and R2 will fail. used for the experiments. G  1  F * 1  F 2 1  2  (i) The traffic pattern

n 1   n 1  The parameters used were as follows –     Type of traffic Constant Bit Rate 1  S1ij * 1  S 2ij i 1   i 1  j i  1 j  i  1     Packet Size 512 bytes For simplicity, we assume that the stability of each link is S. So we have Packet Rate 4 pkts/sec

Slij S, i,j,k Maximum number of 20 Therefore we have the probabilities connections n1 n1 Table1: Traffic pattern G1  1  S  1  S i1 and n 1 n 1 (ii) Scenario description    

G2  1  S  *  1  S  BonnMotion is a Java software which creates and analyses i 1   i 1  mobility scenarios. It is developed within the Communication n 1 n 1 Systems group at the Institute of Computer Science of the University 1  S * 1  S  of Bonn, Germany, where it serves as a tool for the investigation of mobile ad hoc network characteristics. The scenarios can also be 2 2 n1 exported for the network simulators ns-2, ns-3, GloMoSim/QualNet, 1  S  G11  G   COOJA, MiXiM, and ONE. Several mobility models are supported, n1 namely Q1 s 1 In general, for the such parallel routes we would have  the Random Waypoint model, k  the Random Walk model, GGk1   the Gauss-Markov model, So, the probability of communication link breakage between source  the Manhattan Grid model, and Destination reduces exponentially if parallel routes are used. In other words, communication becomes more stable when multiple  the Reference Point Group Mobility model, routes are used.  the Disaster Area model,  the Random Street model, 5. PERFORMANCE EVALUATION OF PREEMPTIVE DSR  and more.[11] IN A BATTLEFIELD SCENARIO

As explained earlier, the Preemptive DSR routing protocol uses a combination of table-driven and reactive methods to achieve optimal performance. It has been found previously, that PDSR achieves a higher packet delivery fraction and lower latency than the table- driven protocols. Further, it also adapts well to node mobility and link changes.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

It generates the movements of nodes in an ad hoc network as a trace n file which can be imported into ns-2..The following metrics were (timePacketRe ceivedi  timePacketSenti ) used to depict a battlefield scenario. AED  i0 totalNumberOfPacketsRe ceived Dimensions 2000*2000 Mobility Model Reference Point Group A higher value of end-to-end delay means that the network is Mobility Model congested and hence the routing protocol doesn’t perform well. The (RPGM) upper bound on the values of end-to-end delay is determined by the No. of nodes 50 application. For example multimedia traffic such as audio and video Min. speed 1 m/s cannot tolerate very high values of end-to-end delay when compared to FTP traffic. [17][20] Max. speed 5 m/s (iv) Research methodology Average number of 10 Three parameters in the battlefield scenario were varied - pause time, nodes in a group the total number of nodes and average number of nodes in a group Probability of group 0.01 and their impact on the three metrics described above were studied. change The results are discussed in the next section.

Pause time 60 sec 6. RESULTS i. Effect of varying the number of nodes The number of nodes was varied from 50 to 100 and the effect on Table 2: Parameters for the battlefield scenario PDF, NRL and AED was studied. The results can be found in table 3

and figures 1, 2 and 3. (iii) Metrics: The following metrics are used for performance evaluation. a. Packet Delivery Fraction (PDF): This is the ratio of total number It is found that the packet delivery fraction decreases as the number of packets successfully received by the destination nodes to the of nodes in the network increases. This is due to the fact that as number of packets sent by the source nodes throughout the number of nodes increases, the congestion in the network also simulation. increases and hence the number of lost packets due to retransmission also increases. Further, since PDSR uses a table driven approach, the numberOf Re ceivedPackets processing delay at the nodes also increases with an increase in the PDF  size of the network thereby accounting for the higher end-to-end numberOfSentPackets delay. The normalized routing load increases with an increase in number of nodes due to an increase in the routing packets in the This estimate gives us an idea of how successful the protocol is in network. delivering packets to the application layer. A high value of PDF indicates that most of the packets are being delivered to the higher layers and is a good indicator of the protocol performance. Table 3: Effect of varying the number of nodes b. Normalized Routing Load (NRL): This is calculated as the ratio between the no. of routing packets transmitted to the number of packets actually received (thus accounting for any dropped packets). numberOfRoutingPacketsSent NRL  numberOfDataPacketsRe ceived

This metric gives an estimate of how efficient a routing protocol is since the number of routing packets sent per data packet gives an idea of how well the protocol maintains the routing information updated. Higher the NRL, higher the overhead of routing packets and consequently lower the efficiency of the protocol.

c. Average end to end delay (AED) : This is defined as the average delay in transmission of a packet between two nodes and is calculated as follows-

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 1: Effect of varying the number of nodes on the pause time Figure 3: Effect of varying the number of nodes on the Normalized Routing Load

The blue circles in figures 1, 2 and 3 represent the “optimal points” which corresponds to the highest PDF, lowest end-to-end delay and the lowest normalized routing load. It is found that for 60 nodes we achieve this optimal point.

ii. Effect of varying the pause time The effect of varying the pause time on the three metrics are shown in table 4 and the corresponding graphs are shown in figures 4,5 and 6. It can be inferred that as pause time varies, the packet delivery fraction also increases. This is due to the fact that as pause time increases, the relative mobility of the nodes decreases, and hence the congestion also decreases in the network. The end-to-end delay also decreases as the pause time is increased. This can be explained as follows – as the pause time increases, the network topology is relatively stable and hence the number of stale routes in the routing tables decreases. Thus route discovery and maintenance take less Figure 2: Effect of varying the number of nodes time. This also reduces the number of routing packets in the network, on the Average end-end delay thereby decreasing the NRL.

Table 4: Effect of varying the pause time

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 4: Effect of varying the pause time on PDF Figure 6: Effect of varying the pause time on NRL

From figures 4, 5 and 6 it can be inferred that for a pause time of 20 sec (represented by a blue circle), we obtain optimal values for the three metrics.

iii. Effect of varying the average number of nodes The effect of varying the average number of nodes on the three metrics is shown in table 5.The graphs for the three metrics are shown in figures 7,8 and 9.

Table 5: Effect of varying the average number of nodes

Figure 5: Effect of varying the pause time on average end to end delay

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

From figure 4.9 it can be inferred that the PDF decreases as the average number of nodes in a group is decreased. This is due to the face that as the average number of nodes increases, the density increases, thereby causing more congestion in the network. Since PDSR uses HELLO messages for neighbor detection, as the node density increases, the number of such packets also increases, thereby decreasing the PDF. The effect of increasing the average number of nodes on the average end-to-end delay is shown in figure 8. It is found that the delay decreases the density increases, thereby indicating that PDSR scales well to the network density. Further by not using source routing, it achieves lower latency due to a lesser packet overhead.

Figure 8 shows the effect of varying the average number of nodes in a group on the routing load. In general, PDSR has less routing overhead achieving a peak load of about 0.32 when the average number of nodes in a group is 9 (represented by blue circles in the graphs). From the graphs, it can be inferred that the optimal point corresponds to 8 nodes per group.

Figure 7: Effect of varying the average number 7. CONCLUSION of nodes on the PDF For the battlefield scenario, PDSR has found to perform well for lower pause times (20 sec), higher density of nodes (9 per group) and smaller networks. As the network size increases, the performance drops due to a table-driven approach. However, since it does not use source routing, it has a much lower end to end delay for In order to analyze the performance of routing protocols in practice, such a scenario-based approach is vital. It also helps identify the suitable routing protocol for an optimal network size, the mobility of the nodes, the network density and a given traffic pattern. A more comprehensive study of other routing protocols such as DSR, TORA, DSDV, etc. is needed to choose the right protocol for a given scenario.

Figure 8: Effect of varying the average number REFERENCES of nodes on the AED [1] Banner R., Orda A.. Multi-path Routing Algorithms for Congestion Minimization. IEEE/ACM Transactions on Networking, vol. 15(2), April 2007, pages 413-424.

[2] Z. Ye, S. V. Krishnamurthy, S. K. Tripathi, "A Framework for Reliable Routing in Mobile Ad Hoc Networks, " Proceedings of IEEE INFOCOM 2003, 30 March 30- 3 April 2003, Vol. 1, pp. 270 - 280.

[3] Barritt, Brian J.; Sheikh, Shaya; Al-Najjar, Camelia; Malakooti, Behnam “Mobile ad hoc network broadcasting: A multi-criteria approach” , International Journal of Communication Systems, Vol. 24, Issue 4, pp. 438-460, April 2011. [4] Safa, H.; Mirza, O. “A load balancing energy efficient clustering algorithm for MANETs”, International Journal of Communication Systems, Vol. 23, Issue 4, pp.463-483, April Figure 9: Effect of varying the average number 2010. of nodes on the NRL

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[5] Stefano Basagni, Marco Conti, Silvia Giordano, Ivan [16] P Ng, S Liew. Throughput Analysis of IEEE 802.11 Stojmenovic, “Mobile Ad Hoc Networking”, ISBN: 0- 471- Multi-hop MANETs Networks. IEEE/ACM Transaction on 37313-3, Wiley-IEEE Press: Chapter 12: Ad hoc networks Networking, June 2007. Security Pietro Michiardi, Refik Molva [17] F. De Rango, P. Lonetti, S. Marano, "MEA-DSR: A [6] Hongmei Deng, Wei Li, and Dharma P. Agrawal, Multipath Energy-aware Routing Protocol for Wireless Ad “Routing Security in Wireless Ad Hoc Network,” IEEE Hoc Networks, " Proceedings of the Seventh Annual Communications Magzine, vol. 40, no. 10, October 2002. Mediterranean Ad Hoc Networking Workshop, Palma de Mallorca, Spain, 25-27 June, 2008, pp. 215-225. [7] Ho, Ai Hua; Ho, Yao Hua; Hua, Kien A. “Handling high mobility in next-generation wireless ad hoc networks”, [18]The Network Simulator NS-2: www.isi.edu/nsnam/ns/. International Journal of Communication Systems, Vol. 23, Issue 9-10, pp. 1078-1092, Sep-Oct 2010. [19] Xiaoxia Huang, Yuguang Fang. Performance Study of Node-Disjoint Multi-path Routing in Vehicular MANETs [8] V.Ramesh, Dr.P.Subbaiah, K.Sangeetha Supriya Networks. IEEE Transactions on Vehicular Technology, vol. “Modified DSR (Preemptive) to reduce link breakage and 58, pages 1942-1950, May 2009. routing overhead for MANET using Proactive Route Maintenance (PRM)” in the Global Journal of Computer [20] Kao, Hui-Hsiang; Wu, Peng-Jung; Lee, Chung-Nan ” Science and Technology, vol.9, issue 5, January 2010, pp 124- Analysis and enhancement of multi-channel MAC protocol for ad 129. hoc networks”, International Journal of Communication Systems, Vol. 24, Issue 3, pp.310-324, March 2011. [9] Krings, A.W., Ma, Z., "Fault-Models in Wireless Communication: Towards Survivable Ad Hoc Networks", [22] C. Siva Ram Murthy, B.S. Manoj, “Ad Hoc Wireless Military Communications Conference, 2006. MILCOM 2006. Networks : Architectures and Protocols”, Prentice Hall IEEE, pp 1 – 7. publishers, May 2004, ISBN 013147023X

[10] Dekar, L., & Kheddouci, H., (2008). A Cluster Based [23] Z. Zhang, G. Dai and D. Mu, Bandwidth-Aware Multi- Mobility Prediction Scheme for Ad Hoc Networks”, Ad Hoc path Routing Protocol for Mobile Ad Hoc Networks. Springer Networks Journal. Vol. 6(2), April 2008, Elsevier. Ubiquitous Intelligence and Computing, Lecture Notes in Computer Science, pages 322-330, 2006. [11] S. Jiang, "An enhanced prediction-based link availability estimation for MANETs," IEEE Transactions on [24] Wu, Chien-Min; Hou, Ting-Chao; Leou, Maw-Lin; Liaw, Yi- Communications, vol. 52, no. 2, pp. 183-186, 2004. Ching; Chan, Ming-Chieh “Adaptive backoff scheme for ad hoc networks based on IEEE 802.11”, International Journal of [12] Chen, Yuh-Shyan; Hsu, Chih-Shun; Chen, Po-Ta “A multiple Communication Systems, Vol. 23, Issue 12, pp.1500-1520, Dec relay-based medium access control protocol in multirate wireless 2010. ad hoc networks with multiple beam” , International Journal of [25] Feng, Wei; Elmirghani, Jaafar M. H. “Lifetime evaluation in Communication Systems, Vol. 23, Issue 5, pp. 596-632, May energy-efficient rectangular ad hoc wireless networks” 2010. International Journal of Communication Systems, Vol. 23, Issue 12, pp.1632-1650, Dec 2010. [13] B. L. Sun, S. C. Pi, C. Gui, et al, "Multiple Constraints QoS Multicast Routing Optimization Algorithm in MANET based on GA, " Progress in Natural Science, Vol. 18, No. 3, 2008, pp. 331-336. Authors’ Brief

[14] B. L. Sun, C. Gui, Q. F. Zhang, H. Chen, "Fuzzy V. Ramesh received his B.Tech from Controller Based QoS Routing Algorithm with a Multiclass N.B.K.R.I.S.T, Vidyanagar, AP in Computer Scheme for MANET, " International Journal of Computers, Science & Engineering and M.Tech in IT Communications & Control, Vol. IV, No. 4, 2009, pp. 427-438. from Sathyabama University, Chennai. Presently he is working as Associate [15] B. L. Sun, C. Gui, Q. F. Zhang, et al, "A Multipath on- Professor in the department of Computer Demand Routing with Path Selection Entropy for Ad Hoc Science & Engineering at Sree Networks, " Proceedings of The 9th International Conference Vidyanikethan Engineering College, for Young Computer Scientists (ICYCS 2008), Zhang Jiajie, Tirupati, AP. He has published several papers in various Hunan, China, 18-21 November, 2008, pp. 558-563. International & National Conferences and Journals. Presently he is pursuing his Ph. D in the field of Ad-hoc networks at Sathyabama University, Chennai. His research interests include Operating Systems, Computer Networks and MANETs.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Dr P. Subbaiah received M.Tech(D.S.C) from JNTU and Ph. D from S.K University, Ananthapur in the area of fault tolerant systems. He has published several papers in international, national conferences and journals. He guided 6 research scholars. Presently he is working as Professor in ICE, Sree Sai Ram Engg. College,West Tambaram, Affiliated to Anna University, Chennai, India. His research interests include Mobile ad-hoc networks, Digital image processing and VLSI design.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Development of a Mobile Feedback System for Health Institutions in Nigeria

S. Okuboyejo, S. Akor & A. Adewumi Department of Computer and Information Sciences Covenant University Ota, Nigeria [email protected], [email protected], [email protected]

ABSTRACT Feedback is very essential in any organizational set up and this does not exclude health institutions. With feedback an institution is able to stay in touch with the needs and expectations of their customers and to also improve on service provision. However, the process of collecting feedback is of importance. After thorough investigation and observation it was discovered that most health institutions in Nigeria did not have the necessary means of getting consistent feedback from their patients and customers. This paper therefore introduces a mobile application that serves as a feedback mechanism between patients and their health institutions.

Keywords- Feedback mechanism, mobile device, health institutions

1. INTRODUCTION 2. EXISTING SYSTEMS The way and manner in which health care services are delivered in any nation goes a long way in determining the well-being of the 2.1 Email Feedback System citizens of that nation. There is need among other things for One interesting approach in this feedback system is that it can be continual innovation in order to foster improved quality, value and evaluated by clients and users. Email is not a new concept, but its patient experiences in health care delivery [Shih et al., 2008]. use for feedback communication is worth considering [Keil & Innovation will come when feedback is taken seriously [Cirillo & Johnson, 2002]. Nicereply is an example of an email feedback Fisher, 2005]. Feedback is a necessity in every organizational system [www.nicereply.com]. This application turns an email into setup and this does not leave out health institutions. Through a customer service feedback system [McCarthy, 2010]. It functions feedback, health institutions will get to identify with their patients’ by inserting a link at the bottom of every mail sent by an experiences and expectations. Being armed with this information, organization to its customers. With this link, the customers can rate they would be better informed on how to improve their services. the response of the organization. It is common to find today persons who possess at least one email address which makes them In recent times, communication is being enhanced through reachable through this channel [Milev, 2010]. The drawback of this electronic means such as emails, text messages, and phone calls. method is spam filter. Oftentimes, emails which get sent from one Mobile devices are also becoming ubiquitous among all and sundry and the same server in short periods of time and to a big mass of especially across the nations of Africa and Nigeria in particular. recipients are marked as spam. Such messages get hidden from the After thorough investigation and observation it was discovered that user or else directly deleted [Milev, 2010]. This may bring about a most health institutions in Nigeria did not have the necessary break in communication. means of getting consistent feedback from their patients and customers. As results of this, many institutions have not given their patients the satisfactory care and services they need in order to 2.2 Telephone-based Feedback System improve their health. Survey Crafter [www.surveycrafter.com] is a well-known example of a telephone-based feedback system. It is software that strives to Research however, shows that by employing electronic means, make writing, administering, and analyzing web, paper and feedback collection could be greatly enhanced [Powney & Hall, telephone-based surveys easy. This however is not a mobile 1998] especially by leveraging on the mobile platform [WHO, feedback system, but rather a “direct approach” method to feedback 2011]. This paper therefore introduces a mobile application that collection. The clients of the system rent international phone lines serves as a feedback mechanism between patients and health to directly call the potential target groups. This brings fast results. institutions. The rest of this paper is structured as follows: section 2 Owing to the direct communication between the companies which reviews the existing systems in this domain, in section 3 the started the feedback process, its consumers can directly tell their methodology adopted in the research is reported while in section 4, opinion on the stated questions and can even place questions on the results are discussed. Section 5 concludes the work and gives their own. This method has two big drawbacks. the scope for future work.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

As a result of the of the rented international phone lines, such method costs a lot of money, which cannot be covered by smaller Patient Phone UI Feedback Form Medical History Database companies. The second drawback is the need for trained professionals in this sphere, who can carry out the feedback on the line. The direct approach to the target group demands a measured Login approach, big confidence and calmness during the talk and instantaneous evaluation of the results [Milev, 2010]. Checks Username and password with database

Validate 2.3 Mobile Health Surveys

This involves the use of mobile devices for health-related data Display patient's profile collection and reporting. A survey conducted by World Health Organization (WHO) shows that the use of mobile devices for health surveys was low across the six WHO regions [WHO, 2011]. Fill feedback form However, among the six regions, the Americas had the highest with 42%. This is closely followed by the African Region with 31%. Send to database

The two regions reported the highest proportion of Member States Validate with this mHealth initiative [WHO, 2011]. With the increased ubiquity of mobile devices into the African continent, the use of Display Confirmation Message mobile health surveys as a means for collecting feedback can only be on the increase. Fill Medical History Form 3. Methodology Send to database

3.1 Design Architecture Validate

The mobile feedback system presented in this paper was modeled using the Unified Modeling Language. It is structured as a 3-tier application. This consists of a client, middleware and database Display Confirmation Message server. The client refers to mobile devices with which users send feedback to their health institution. The middleware serves as an intermediary between the client and the database server. As an intermediary, it receives the data from the client and forwards it to the database server of the health institution where it can be promptly acted on.

Middle Ware Client/Front-end

Fig. 2: Sequence Diagram for a Patient’s feedback

transmission

A sequence diagram was used to show the flow of feedback from User Mobile Phone users to their health institutions. A class diagram on the other hand was used to show all the entities involved in the feedback system and the relationship between them.

Database Server

Application and web server

Fig 1: Architecture of the mobile feedback system

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The major advantage of the mobile feedback system is that patient, Feedback form health workers, and various governmental and non-governmental -pidn : string organizations can contribute to the effective management of any -full_name : string -medical_insurance : health institution. Also the system serves as a platform in which string -first_visit : string provision of ideas especially from patients and medical * * -fallen_sick professionals can be made. -diagnosis -Fills -affordable -prescribed_med -symptoms -admitted -admission_services -satisfactory_service -comment

* Admin -Userame : string -Password : string 1 +Add User() 1 +Edit User() * Feedback System Database +Update Information() -Patient_medical_history : string +Delete() -patient_record : string Patient -admin : string -User Name : string -feedback_form : string -Password : string +store patient_record() 1 -PIDN : Integer +store admin data() +Fill feedback forms() +store form data() -Analyses 1 +Register() +store feedback form record() +fill_medical history form() +store medical history record()

1 Records officer * -Fills -Username : string -Password : string +Analyse feedback form() * +Analyse medical_history() +print report() Medical history -pidn : string -Analyses 1 -full_name : string -marital_status : string -have_children : string -no_children -next_kin -relationship 1 -asthmatic * -heart_diseases -diabetic -have_cancer -liver_complications

Fig. 3: Class Diagram for the Mobile Feedback System

3.2 Development Tools The system was developed as using open source tools namely: PHP, MySQL, Hypertext Markup Language (HTML), and Android SDK. HTML was used to develop the user interface of the application and PHP was used as the server-side scripting language that allowed for communication between the user interface and database. MySQL served as the database. The prototype was tested out on an Android emulator.

4. DISCUSSION There has been great need over the years for the automation of the activities of many health institutions in Nigeria to facilitate work effectiveness, efficiency, reduce cost and increase accountability. Fig. 4: Screenshot of the Mobile Feedback Form There has also been an increase in the need for patients to express their views on the operations of their health institutions. With the adoption of the mobile feedback system, patients are assured of a smooth running and more effective management of various health institutions.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[4] Milev, M. (2010) Fast, User-friendly Feedback System for the Mobile, B.Sc Thesis [5] Powney, J. and Hall, S. (1998) “Closing the loop: The impact of student feedback on students’ subsequent learning,” The Scottish Council for Research in Education [6] Shih, A., Davis, K.,Schoenbaum, S. C., Gautthier, A., Nuzum, R., and McCarthy, D. (2008) “Organizing the U.S. health care delivery system for high performance,” The Common Wealth Fund [7] World Health Organisation (2011) mHealth: New horizons for health through mobile technologies.

Authors’ Brief

Okuboyejo, Senanu Rita holds a M.Sc in Management Information Systems (MIS). Her research areas include Health Informatics, Technology Adoption and Acceptance in Sub-Saharan Africa, Technology Diffusion in Healthcare, Data mining, Data modeling, and Information management, System Analysis and Design, Database Management. She is currently pursuing a PhD in the area of mobile health adoption and diffusion in Covenant University, Ota, Ogun state, Nigeria. Fig 5: The Feedback Analysis Page

Akor, Samuel graduated in 2011 from the 5. CONCLUSION Department of Computer and Information In order to ensure that the Mobile Feedback System is used to Sciences, Covenant University. He studied maximum capacity, it is made open for further reviews and Computer Science and is currently doing his enhancements. Some features and functionalities were not fully 1-year National Youth Service (NYSC) in implemented. Therefore, there is still room for further research. Abuja. He can be reached at Some suggestions for further research are: [email protected].

 Implementation of more enhanced security measures and control

 The integration of Short Message Service (SMS) module into the system to enable patients communicate with the Adewumi, Adewole Oluwasegun holds a system via SMS. Masters degree in Computer Science from the Department of Computer and Information Sciences of Covenant REFERENCES University. Hi s current research interest is developing cross-platform mobile

applications for healthcare and education. [1] Cirillo, A. and Fisher, A. (2005) “The importance of

continuous customer feedback … and how to get it quickly,”

Strategic Health Care Marketing

[2] Keil, M., and Johnson, R. D. (2002) “Feedback Channels: Using Social Presence Theory to Compare Voice Mail to E- mail,” Journal of Information Systems Education, vol. 13, no. 4, pp. 295-302 [3] McCarthy, B. (2010) “Nicereply turns your email into a

customer service feedback system”, TNW Apps Blog,

Accessible at:

http://thenextweb.com/apps/2010/08/04/nicereply-turns-your-

email-into-a-customer-service-feedback-system/Date

accessed: 19th July, 2012

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Towards Developing an Online Social Media-based Mobile Learning System

1N.A. Ikhu-Omoregbe, 2C.K. Ayo, 3A.A. Azeta & 4V. Macus Department of Computer & Information Sciences Covenant University Ota Nigeria [email protected], [email protected] [email protected], [email protected]

ABSTRACT The advancement of Information and Communication Technology (ICT) and the Internet revolution gave rise to the several learning technologies on the web and mobile platform. During the last decade, the social media network became available for users to socialise and collaborate among peer group. Hence, The integration of e-learning and social media using mobile device as access point is to allow for learning and collaboration anytime, anywhere. This study seeks to provide learning on the social network platform for users to view the application on a mobile device and also foster collaboration among scholars. The system was developed using an open source Content Management System (CMS) Wordpress and Buddypress running on a WAMP or XAMPP server. MySQL was used as database. The usability of the System on the different mobile devices used was evaluated by identifying the usability attributes; designing a questionnaire based on those attributes and then analyzing the results with Statistical Package for Social Science (SPSS). The results showed that the learning system had a good usability score on mobile devices. .

Keywords- Collaboration, Mobile Learning, Social Media, Usability, VoiceXML

1. INTRODUCTION social network such as Facebook, Twitter and LinkdIn is on the increase most especially among young individuals, the crop of The recent advances in mobile technology are changing the which are students of tertiary institutions. The growth and primary purpose of mobile devices from making or receiving popularity of online social networks has created a new world of calls to retrieving the latest information on any subject. In collaboration and communication. More than a billion computer science, mobile computing is mainly about increasing individuals around the world are connected and networked the capability to physically move computing tools and services together to create, collaborate, and contribute their knowledge around [1]. Mobility offers the ability to engage learners of all and wisdom. Despite the importance of online social networks, ages anywhere, anytime [2]. New mobile technology, such as there is relatively little theory-driven empirical research hand-held cellular based devices is playing a major role in available to address this new type of communication and redefining how we receive information. One issue is crystal interaction phenomena [5]. Social networking websites are clear and that is mobile learning is not just about learning using virtual communities which allow people to connect and interact portable devices, but learning among peer and social groups of with each other on a particular subject or to just ‘‘hang out” people [3]. together online.

Advances in computer and communication technologies has Students are heavily immersed in Web 2.0 technologies (i.e. resulted in the development of portable digital devices such as Facebook, twitter, podcasts, wikis, blogs, chats, virtual worlds, cell phones, personal digital assistants, netbooks, iPods, video video sharing and photo sharing). They are crafting on-line cameras, Moving Picture Expert Group3 (MPEG3) players, niches for themselves that seamlessly blend with their off-line Global Positioning System (GPS), and portable e-books for world). Indeed, the Internet is playing an increasingly important enhanced participation in online communities of learners. role in not only students’ social life, but also academic [6, 7]. Statistics has shown that 4.7 billion mobile cellular subscriptions Educators are now turning to Web 2.0 tools, drawing upon their exist globally in 2009 [4]. The pedagogical application of these ability to assist in creating, collaborating on and sharing content. devices has lead to the development of ‘Mobile Learning’, a As a result of this, the usage of social sharing sites is increasing rapidly expanding area of technology supported learning. daily [8]. Learners have indicated the need to use portable devices to learn Wireless devices are highly individualized with collaborative on motion. As the most important social technology used communications facility. This advancement give faculty flexible worldwide, mobile device plays an important role in education. tools for complementing the existing technologies and extending the learning beyond the classrooms and homes from remote A social media network platform is one that provides a medium places like airports or trains where students do not have access for interaction by groups of people making it easy to share to computers and the Internet [9]. A learning technology which information (such as lecture materials, pictures and ideas) across is mostly used by the visually impaired learners is Voice-based a circle of people or groups. Statistics has shown that the use of e-learning application.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Voice-based social network is used to search for ad-hoc People use the social media for several other activities including: information, documentation and sharing of images and video, communicating, collaborating, seeking expert advice, sharing and access to social networking sites. VoiceXML-based mobile multimedia, presenting opinions, sharing reviews, entertainment, application allows users to connect to a Internet or Intranet Public health [17], Tax Administration[18], Insurance [19] and server by simply dialing a telephone number for a mobile phone. business growth [20]. An increasing number of people are using VoiceXML is also known as VXML. It is one of the tools for social media in their buying decisions. It is so because social developing voice-enabled e-learning applications. It is a web- media helps them filter the large amount of information based markup language for representing human-computer available by being able to rely on comments from their friends dialogs, just like the HyperText Markup Language (HTML). But and like-minded individuals [19]. In terms of outreach, social while HTML assumes a graphical web browser, with display, media allows individuals, companies, organizations, keyboard and mouse, VoiceXML assumes a voice browser with governments, and parliamentarians, to reach large numbers of audio output (computer-synthesized and/or recorded), and audio people[22]. input (voice and/or keypad tones) [10]. VoiceXML technology allows a user to interact with the Internet through voice- Recent technological advancement has proved that learning has recognition technology by using a voice browser and/or the moved from being web-based to mobile learning platform. telephone. The major goal of VoiceXML is to bring the Reason being that the Internet service for web-based technology advantage of web-based development and content delivery to may not always be available everywhere, every time, on real Interactive Voice Response (IVR) system [11]. time basis. With this in mind, it becomes questionable to only rely on web-based learning to provide all the required social There are several social media technologies that promote e- learning needs. In the field of behavioural psychology, social learning. They include: Edublogs Campus, Elgg, Google learning is defined as the kind of learning by individuals that collaborative tools. They offer the potential to encourage happens through observation or interaction with their social collaboration; enable user-generated content or input; provide context [23]. One of the key attributes of social learning is effective way to share resources; and facilitate informal or collaboration. Social learning uses behavioural and formal learning [12]. There exist several social e-learning observational learning technique to attain collaboration. It is systems such as Moodle, Sakai, Claroline, Ilias, Cramster, believed that the behaviour of a learner is influenced by Cloudworks, Mixable amongst others. observing other learners among peer groups.

The main contribution in this study is to show how the 3. SYSTEMS DESIGN AND ARCHTECTURE convergence of social network application and mobile learning has enhanced the accessibility of e-learning system. The The design of the system as presented in this section contains a objective of the study is to provide access to e-learning content process model and architecture of the system. The process across two different platforms to allow for effective model of the system shown in Figure 1 contains the sequence of collaboration of peer groups in e-learning. activity of the system. The symbols used in the model are Business Process Modeling (BPM) notations which are also 2. RELATED LITERATURE close to flowchart symbols. In Figure 1, a rectangle represented an activity, rhombus as decision, rectangle with round edges as Social networking is built on the idea of interaction and sharing. action, circle as event and directed arrow as sequence of process However, such information sharing and collaboration has some flow. A mobile user login into the system using a mobile phone teething problems such as privacy issues and integrity of friends via graphical user interface (GUI). The login profile of the user on social networking sites. It may also cause health challenges is authenticated and the service type is determined. If service as a result of staying too long sitting in one place browsing the type is Group, then Courseware is pulled for the user. If forum is Internet. This tends to affect the operation of genes in the body selected, then comments are ready to be posted. If members are system. Additionally, social media interaction when used for selected, then chat/messages services are available. The learning does not give sufficient room for explanation and architecture of the system shown in Figure 2 was drawn using clarification [13]. It is no gain saying the fact that most youths schematic architectural design. use mobile device to access social network sites on the Internet. Despite the high popularity of personal use of online social media, a low percentage of students and instructors use them for educational purposes [14]. However, some online social media- based learning resources have been reported in [15] and [14].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 2: The Architecture of the System

The system was developed using an open source Content Management System (CMS) Wordpress and Buddypress running on a WAMP or XAMPP server. JavaScript and AJAX (the scripting language which helps to detect which device is used to access the application whether laptop or mobile). MySQL will be used to manage the database of the application. The usability of the System on the different mobile devices used was evaluated by identifying the usability attributes; designing a questionnaire based on those attributes and then analyzing the results with SPSS software. The results showed that overall the learning system had a good usability score on the mobile devices used.

4. SYSTEMS IMPLEMENTATION

Figure 1: Process Model of the System The Welcome Page (see Figure 3) shows where a user gets to after logging into the System. It gives a brief history of the users’ personal activity on the system. It also allows a user start The description of the architecture in Figure 2 is presented as getting used to the system by posting comments, uploading follows: The learner connects with the logic layer using a pictures and other activities that shows the user is active on the smartphones. The access path for the application is through System. HyperText Transmission Protocol (HTTP).

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 4. The Profile Menu Page on Blackberry

The usability of the System on the different mobile devices used was evaluated by identifying the usability attributes; designing a questionnaire based on those attributes and then analyzing the results with SPSS. The results showed that overall the learning system had a good usability score on the mobile devices used. Figure 3: Welcome page on iPod 5. SYSTEMS EVALUATION The Profile Menu page (see Figure 4) consists of Activity, Profile, Messages, Friends, Groups, Forums, and Settings. Each To test the performance of the system users were told to use of these links shows users personal activities and gives general their mobile device to access the system in order to observe its information of users profile. The Profile menu is located as a performance i.e which features were accessible on the devices. drop-down link under the main menu on top right corner of the After testing the System’s functionality on the mobile platforms, welcome page (for the Blackberry phone) and a direct click on the findings are summarised in Table 3: the Profile link for the iPad.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Table 1: Experiment Findings System Functionality iPad iPod Touch Blackberry Phone Android Nokia Tablet/Phone Phone Create Account      Login      Add Friend      View Member Profile and     x location Create Forum   x   Make comments to existing      discussions Chat/Messages      Join Groups      Courseware      Edit Profile      Save Profile      View System Activity      Logout     

According to the ISO 9241-11 standard, usability refers to “the 6. CONCLUSION AND FUTURE RESEARCH extent to which a product can be used by specified users to achieve specified goals with effectiveness, efficiency and satisfaction in a With the on-line social media-based mobile learning system specified context of use”. In evaluating this System, the following provided in this study, scholars and instructors can now have a usability factors were considered: Attractiveness/ Interestingness, platform where effective teaching and learning can take place. The Simplicity, Browserbility, Navigability, Completeness, system will allow for collaboration and interaction because of the Interactivity and satisfaction. A total of 26 people participated in integration of the social media concept which is a tool that is the usability study. According to [21], this is a suitable number widely used especially by young scholars. Furthermore, due to the required for such usability study. The table gives the descriptive increasing trend towards development and usage of smartphones, analysis of the data gathered from the questionnaire. this System can be accessed by mobile devices bringing knowledge closer to learners and enhancing information sharing at Table 2: Descriptive Statistical Analysis of Questionnaire Data any given time. Usability Mean Standard Variance Attributes Rating Deviation Areas that require further research include Ethical, legal and Attractiveness 4.38 0.590 0.348 privacy issues and a number of pedagogical limitations affecting e- Simplicity 4.43 0.598 0.357 learning and social media [4]. The impact of social media usage Browserbility 4.43 0.507 0.257 habits on the effectiveness of e-learning platforms has not been Navigability 4.19 0.602 0.362 examined yet especially in the light of cultural differences [23]. Completeness 4.05 0.805 0.648 Social learning theory and analysis will also be considered in the Interactivity 4.19 0.512 0.262 system implementation and evaluation. Satisfaction 4.19 0.680 0.462 REFERENCES

Numerous usability studies suggest that a System with ‘Good [1] Gilliot J.M., Garlartti S. Rebai I. Pham nguyen C. (2012), “A Usability’ should have a mean rating of 4 on a 1-5 scale and 5.6 on Mobile Learning Scenario improvement for HST Inquiry Based a 1-7 scale [22]. The 1 – 5 scale approach was used for testing the learning”. usability of this work. We can therefore conclude that this System has ‘Good Usability’ based on the mean ratings of the usability Online at: /www2012.wwwconference.org/proceedings/ attributes, shown in Table 4 above. nocompanion/EWFE2012_001.pdf

[2] O’Malley C., Vavoula G., Glew J.P., Taylor J., Sharples M.,

Lefrere P. (2009): Guidelines for learning/teaching/tutoring in a

mobile environment, MOBIlearn/UoN,UoB,OU/D4.1/1.0, pp. 6.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[3] Sharples M, Milrad M, Arnedillo Sánchez I, Vavoula G(2009): [15] Smith M. & Berge Z. (2009), “Social Learning Theory in Mobile Learning: Small devices, Big Issues.Edited by: Balacheff Second Life”, MERLOT Journal of Online Learning and Teaching. N, Ludvigsen S, de Jong T, Lazonder A, Barnes S. Technology Vol. 5, No. 2, June 2009. Enhanced Learning: Principles and Products. Heidelberg: Springer; 2009:233-249. [16] Taxpayer(2011),”Social Media Technologies and Tax Administration, Forum on Tax Administration: TaxPayer Services [4] Pimmer C. (2012), How mobile learning and social media can Sub-Group. support learners and health professionals in “low resource settings”, University of Applied Sciences and Arts, Northwestern [17] SocialMedia WG (2012), “The Use of Social Media in Switzerland. WSIS forum 2012, 14-18 May Geneza. Insurance”. Social Media (D) Working Group of the Market Regulation and Consumer Affairs (D) Committee Adopted [5]Cristy M. K., Pui-Yee C. Matthew K. O. Lee(2011), December 20, 2011. National Association of Insurance “Online social networks: Why do students use Facebook?”, Commissioners Elsevier, Computers in Human Behavior Volume 27, Issue 4, July 2011, Pages 1337–1343 [18] Stelzner M. A. (2012), “Social Media Marketing Industry Report, How Marketers Are Using Social Media to Grow Their [6] Petrović N., Petrović D. and jeremić V.(2012), “Possible Businesses”. April 2012. Sponsored by Social media examiners. Educational Use of Facebook in Higher Environmental Education”, ICICTE 2012 Proceedings. pp 355-362. [19] Dewing M. (2010), Social media Library of parliament, background paper, Publication No. 2010-03-E. Ottawa, Canada, [7] Lego Muñoz, C., Towner, T. (2009). Opening Facebook: How Library of Parliament. to Use Facebook in the College Classroom. In I. Gibson et al. (Eds.), Proceedings of Society for Information Technology & [20] Kilvington M. (2007), “Social learning as a framework for Teacher Education International Conference 2009, pp. 2623- building capacity to work on complex environmental management 627. Chesapeake, VA: AACE. problems”. Work supported by the FRST-funded Building Capacity and Research programme. November, 2007. [8] Ktoridou D. and Stavrides L., (2012), “Facebook - A Social Networking Tool for Educational Purposes: Developing Special [21] Faulkner, L. (2003): Beyond the five-user assumption: Interest Groups”, ICICTE 2012 Proceedings. pp 363-375. Benefits of increased sample sizes in usability testing, Behavior Research Methods, Instruments & Computers, 35(3), 379 – 383 [9 ]Motiwalla L. F. (2005): Mobile learning: A framework and evaluation, ScienceDirect, Computers Education vol. 49 pp. 581– [22] Sauro, J. and Kindlund, E. (2005): A Method to Standardize 596 Usability Metrics into a Single Score, CM, CHI, April 2-7, Portland, Oregon, USA. [10]Gallivan P., Hong Q., Jordan L., Li E., Mathew G., Mulyani Y., Visokey P. and Tappert C., (2002), “VoiceXML Absentee [23] Scheel A. (2012), “Social Media and e-learning: Linking System, Proceedings of MASPLAS'02. The Mid-Atlantic Student usage habits of Social Media Tools and the effectiveness of e- Workshop On Programming Languages and Systems Pace learning in China and Germany. E-Learder Conference April 2012. University. Retrived online 10th January 2010 from http://csis.pace.edu/csis/masplas/p10.pdf Authors Brief [11] Voiceportalwhite paper(2001), available online at : http://www.medialab.sonera.fi/workspace/VoicePortals.pdf accessed 21st March 2008-03-21 Dr. Ikhu-Omoregbe, Nicholas has a B.Sc [12] Ayo C. K, Mbarika V, Ukrakpor U, Emebo O (2011): degree in Computer science from the Implementing a Social Media-Based e-Learning System: Nigeria a University of Benin, Benin city, an M.Sc case Study, pp. 4. degree in Computer Sciences from the University of Lagos, and a PhD degree in [13] Zaidieh A. J. Y. (2012), “The Use of Social Networking in Computer Science from Covenant Education: Challenges and Opportunities”, World of Computer University, Ota, Nigeria. His research interests include: Software Science and Information Technology Journal (WCSIT) ISSN: Engineering, Mobile Computing, Mobile Healthcare and 2221-0741 Vol. 2, No. 1, 18-21, 2012 Telemedicine Systems, and Soft Computing. He currently lectures at Covenant University. He is a member of the Institution of [14] Chen B. and Bryer T. (2012), “Investigating Instructional Electrical and Electronics Engineers Strategies for Using Social Media in Formal and Informal Learning”, Vol 13 No1. The International Review of Research in Open and Distance Learning

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Prof Charles K. Ayo is the Vice Chancellor Covenant University, Ota, Nigeria. He holds the B.Sc., M.Sc. and Ph.D degrees in Computer Science. His research interests include mobile computing, Internet programming, e- business and government, and object oriented design and development. He is a member of the Nigerian Computer Society (NCS), and Computer Professional Registration Council of Nigeria (CPN). He is a Professor of Computer Science and MIS, Covenant University, Ota, Ogun State, Nigeria.

Dr. Azeta, A. Ambrose is a lecturer in the Department of Computer and Information Sciences, Covenant University, Ota, Nigeria. He holds B.Sc., M.Sc. and Ph.D in Computer Science from University of Benin, University of Lagos and Covenant University respectively. His current research interests are in the following areas: Software Engineering, Algorithm Design and Mobile Computing. He currently lectures at Covenant University. He is a member of the Nigerian Computer Society (NCS) and Computer Professional Registration Council of Nigeria (CPN).

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Enhancement of Console-based Dumpbin System to Display API Functions

S.C. Chiemeke & O.E. Osaghae Department of Computer Science University of Benin Benin City, Nigeria. [email protected], [email protected]

ABSTRACT Program analysis is a reliable process of revealing the intent of an executable program to determine whether it is malicious or not. Microsoft Dumpbin software is a program analysis tool that can disassemble executable program. Although this tool can effectively disassemble executable code of a program, revealing its interaction with the machine of implementation, but it has some restrictions. The first restriction is that it was developed mainly for console use. The second is that it only displays memory pointer address of a system function instead of its function name. In this paper, we proposed how to modify Microsoft Dumpbin software to replace its memory pointer addresses, with their equivalent Application Programming Interface (API) names. This modification will help antivirus developers to add Dumpbin software as a plug-in to their antivirus engines instead of using it as just a console analysis tool.

Keywords- Assembly Language, Malicious Software and Reverse Engineering

1. INTRODUCTION

The term reverse engineering describes the process of analyzing Examples of disassembler tools are OllyDbg; a powerful a hardware or software product in order to gain insight into their disassembler that has the strength of providing powerful inner functionality. The field of malicious software analysis is code-analysis features. WinDbg; a disassembler that is divided into three groups: static analysis, dynamic analysis and integrated with Windows operating system and it is provided malicious code de-obfuscation. Static analysis of a malicious free of charge to users. Interactive Disassembler (IDA) Pro; code is a form of reverse engineering whereby an analyst is an extremely powerful disassembler that supports a variety attempts to determine the full functionality of a malware sample of processor architectures and is capable of producing a without actually running the code under study [2]. However, it is powerful flowchart for a program’s function [3] also a way to understand the mechanisms of malicious software, and the prerequisite for automated semantic virus detection tools. Windows Applications use the Application Programming This process is significantly more time consuming and requires Interface (API) rather than making direct system calls. The advanced knowledge of programming languages (most API consists of several important Dynamic Link Libraries commonly C and C++). Dynamic analysis of malicious codes (DLLs). For each DLL, the corresponding APIs can be refers to the technique of monitoring a program’s interaction with referenced either by name or by ordinal number. An the system while it is running on an emulated environment [2] executable program which uses the Windows API is called a and [10]. Portable Executable (PE) file. This PE file has a format which includes an import section that determines which Executable dumping is an important step in reverse engineering DLLs are imported by the executable file [12], [11] and [9]. because it tries to expose what an executable program does. An example of executable dumping tool is Dumpbin; a Microsoft’s console-mode tool for dumping aspects of portable executable 2. RESEARCH OBJECTIVES files. Dumpbin can dump a module’s import and export directories, relocation tables and symbol information [3]. The objectives of this research are as follows: Disassembler is one of the most important reverse engineering 1. To modify Microsoft Dumpbin so that memory tools which decode binary machine code into a readable pointer addresses are replaced with equivalent API assembly language text. This process is somewhat similar to functions names. what takes place within a CPU while a program is running. 2. To proposed a procedure of how the modified Therefore, a disassembler is platform-specific, although there are Dumpbin will be implemented. disassemblers that contain specific support for more than one platform. The problem with a disassembler is its inability to produce an output that is as close as possible to an actual assembler source file that would compile to the same executable again. Moreover, it aids analysis by also processing information other than in the code sections, notably the import section which contains all system and library calls used by the program [6].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

3. MICROSOFT DUMPBIN TOOL AS A The ebp register is generic and mostly used as the stack base DISASSEMBLER pointer. The base pointer usually points to stack position right after the return address for the current function. The register Dumpbin software tool is a command-line utility supplied esp is a processor’s stack pointer. The stack pointer stores the with Microsoft Visual Studio suite of tools. Dumpbin helps to current position in the stack, so that any thing pushed to the examine the contents of a Windows Portable Executable (PE) stack gets pushed below the address, and this register is file. It is capable of displaying a wide range of information updated accordingly [3] and [5]. The mov instruction takes related to Windows Portable Executable (PE) files. The two operands: a destination operand and a source operand, options used to operate Dumpbin offer the ability to extract and simply moves data from the source to the destination. information from sections of a PE binary, including symbols, Arithmetic operations in assembly language are addition imported function names, exported function names and (add), subtraction (sub), multiplication (mul) and division disassembled code [5] and [1]. (div). Operands are compared using the cmp instruction, which takes two operands. The operand cmp records the The format for typing Dumpbin is as follows: dumpbin result of the comparison in the processor’s flag. Function . The command-line calls are implemented using two basic instructions in parameter is the name of the program that is about to be assembly language. The call instruction invokes a function examined. The parameter is a set of optional and the ret instruction returns to the caller. The call command-line arguments that specify the type of information instruction pushes the current instruction pointer onto the about the program to be displayed. The option begins with a stack and jumps to the specified address [3]. slash (/) symbol. For example, /disasm command-line option is used to disassemble a PE file and the assembly code is Currently, Microsoft Dumpbin tool cannot display API displayed as output result of Dumpbin. Another important function names instead, it displays memory pointer addresses example is the /all option, which is also used to disassemble a in square brackets. Another problem with Dumpbin is that is PE file. What make /all option different from /disasm option a console tool and was not primarily designed to be used as is that, apart from disassembly of PE file, it also list all the program plug-in. The purpose of this research is to modifying library routines a PE calls [4]. Microsoft Dumpbin tool so that it can accurately display an equivalent of a disassembled API function of its memory Segments are memory location areas where different sections address pointer. of a PE file are stored during program execution. Some popular segments of PE displayed by Dumpbin are .text, 4. METHODOLOGY .data, .rdata, .bss and .reloc. The .text segment contains the executable instructions and it has READ and EXECUTE In this paper we proposed a modified Microsoft Dumpbin permission. The .data segment contains the initialized global, software tool that has a feature of having its memory pointer static variables and their values. It usually has READ and addresses replaced with API function names. We started the WRITE permissions. The .rdata segment is sometimes called research by disassembling a PE file using Microsoft read-only-data and it contains constants and string literals. Dumpbin tool. We disassembled the PE file, so that we can The .bss segment holds variables that do not have any value display structures of disassembled codes and header file until the program is executed. The .reloc segment stores the section. The portion of the disassemble code of interest are information required for relocating the image of a PE while machine code, assembly code portion, direct call of FF 15 loading in memory [7]. instruction, direct call of E8 instruction, target of direct call of E8 instruction and indirect call of FF 15. We noticed that Assembly language is the language of reverse engineering. there are two types of file header constructs in a PE file. The The binary code generated by assembler program can then be similarities of interest among the type of file headers (file decoded by a computer processor. 32-bit registers have eight header 1 and 2) noticed are machine information, section generic registers: eax, ebx, ecx, edx, esi, edi, ebp and esp. The header and summary information. The differences of interest registers eax, ebx and ecx are generic registers that can be among the type of file headers noticed are values assigned to used for any integer, Boolean, logical, or memory operation. machine information, number of sections, section where The ecx register is generic and is sometimes used as a counter Dynamic Link Library (DLL) and API function names are by repetitive instruction that requires counting. The esi and located. edi are generic registers and they are frequently used as source and destination pointers in instructions that copy Using the structure of disassembled code and file headers memory (si stands for source index and di stands for contents, we labeled the areas of interest in the contents. The destination index). labeling of the contents is to act as tags which will be used to aid modification of the Dumpbin tool. We then gave a summarized procedure for modifying Microsoft Dumpbin disassembler.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

This summarized procedure is a description of how to program and predict its behaviour (whether it has malicious navigate through the PE disassembled contents using the intention or not). In this research, Microsoft Dumpbin labeling tags. Navigating through PE disassembled contents, software was used by typing the executable program’s name gives the idea of how to replace a memory pointer address in at system console prompt. The Dumpbin software a disassembled content with its equivalent API function disassembles the executable program into its equivalent name, in either of the two file headers contents. assembly codes and then displays these codes on the screen or output file. The reason why Dumpbin software was used is 4.1 Structure of a Disassembled Program When a portable because it displays the details of computer resources used by executable program is disassembled, its functionality is a program running in memory of Windows operating system. revealed and this is where the power of reversed engineering Section 4.1.1 shows an illustration of the contents of a lies. This power is in terms of showing the behaviour of the disassembled portable executable program revealed by program to determining whether it has a malicious intention Dumpbin Disassembler tool. Sections 4.1.2 and 4.1.3 shows or not. The process of disassembly is the first step taken by the illustrations of the contents of file header of a portable antivirus experts to reveal the contents of a executable executable program.

4.1.1 Using Dumpbin to Disassembly Portable Executable File Code This section shows a sample of a disassembled executable program using Microsoft Dumpbin tool. For simplicity sake and ease of subsequence references of the codes, each line of the code is numbered and case labels are attached to some portion of the code. The illustration of a disassembled executable program is presented below.

1) Dump of file filename.extension 2) File Type: EXECUTABLE IMAGE 3) SECTION HEADER #1 4) .text name 5) …. 6) Execute Read 7) 8) RAW DATA #1 9) a1a2a3a4a5000 : h1h2 h3h4 h5h6 h7h8 h9h10 assembly code1 10) a1a2a3a4a5a6a7a8: h1h2 h3h4 h5h6 h7h8 h9h10 assembly code2 11) … 12) 13) a1a2a3a4a5a6a7a8: 6A 00 push 0 14) a1a2a3a4a5a6a7a8: A1 00 80 40 00 mov eax, [00408000] 15) a1a2a3a4a5a6a7a8: 68 80 00 00 00 push 80h 16) a1a2a3a4a5a6a7a8: 50 push eax 17) a1a2a3a4a5a6a7a8: 6A 00 push 0 18) a1a2a3a4a5a6a7a8: 6A 01 push 1 19) a1a2a3a4a5a6a7a8: 68 00 00 00 C0 push 0C000000Ch 20) a1a2a3a4a5a6a7a8: 56 push esi case 1 21) a1a2a3a4a5a6a7a8: FF 15 0C E2 40 00 call dword ptr ds: [a1a2a3a4a5a6a7a8h] 22) … 23) a1a2a3a4a5a6a7a8: 6A 00 push 0 24) a1a2a3a4a5a6a7a8: A1 00 80 40 00 mov eax, [00408000] 25) a1a2a3a4a5a6a7a8: 68 80 00 00 00 push 80h 26) a1a2a3a4a5a6a7a8: 50 push eax 27) a1a2a3a4a5a6a7a8: E8 77 B5 5C E7 call 0040E20Ch 28) a1a2a3a4a5a6a7a8: 83 C4 04 add esp, 4 case 3 29) a1a2a3a4a5a6a7a8: 50 push eax 30) a1a2a3a4a5a6a7a8: 6A 00 push 0 31) a1a2a3a4a5a6a7a8: 6A 01 push 1 32) a1a2a3a4a5a6a7a8: 68 00 00 00 C0 push 0C000000Ch 33) a1a2a3a4a5a6a7a8: 56 push esi case 1 34) a1a2a3a4a5a6a7a8: FF 15 0C E2 40 00 call dword ptr ds: [0040E20Ch] 35) … 36) a1a2a3a4a5a6a7a8: 55 push ebp

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

37) a1a2a3a4a5a6a7a8: 8B EC mov ebp, esp 38) … 39) a1a2a3a4a5a6a7a8: 8B E5 mov esp, ebp 40) a1a2a3a4a5a6a7a8: 5D pop ebp 41) a1a2a3a4a5a6a7a8: C3 ret 42) … case 2 43) a1a2a3a4a5a6a7a8: 8B 35 BD 10 00 01 mov esi, dword ptr ds: [01001030h] 44) … case 2 45) a1a2a3a4a5a6a7a8: 8B 3D AC 10 00 01 mov edi, dword ptr ds:[010010ACh] 46) a1a2a3a4a5a6a7a8: 50 push eax 47) a1a2a3a4a5a6a7a8: FF D7 call edi 48) a1a2a3a4a5a6a7a8: 6A 01 push 1 49) a1a2a3a4a5a6a7a8: 6A 00 push 0 50) a1a2a3a4a5a6a7a8: 53 push ebx 51) a1a2a3a4a5a6a7a8: FF D6 call esi 52) … case 4 53) 0040E20C: FF 25 0C 40 00 jmp dword ptr ds:[0040E204h] 54) … 55) Summary 56) 2000 .bss 57) 3000 .data 58) 1000 .idata 59) 1000 .rdata 60) 1000 .reloc 61) 1000 .rsrc 62) 7000 .text

(a) Machine Information: The machine information machine (4-bytes), certain measure must be taken to contains file name and its extension. In addition, it avoid the stack from crashing. For instance, when a contains one section, raw data, and summary programmer wants 2 push arguments to be skipped, the information. The raw data portion contains the usual statement add esp, 8 is used. Again, when it is 3 push memory addresses. This is followed by the semicolon arguments, it is add esp, 0C. The generalized form is character, hexadecimal numbers representation of the when j push arguments wants to be skipped, then the assembly codes and the assembly codes itself. The statement is add esp, 4*j (the result should be in structure of the hexadecimal numbers is the machine hexadecimal form). The add esp, 4, gives direction to code equivalent of the assembly code. This machine the call function statement on line 34 to skip 1 argument information can be seen in line 1-6. (i.e., the push argument on line 26, when the function is (b) (b) Assembly code portion: The assembly code section loading its arguments). of interest in this research is as follows: (ii) Direct call of E8: The hexadecimal code “E8” is (i) Direct call of FF 15: The syntax of the hexadecimal another example of an assembly call statement, and the number “FF 15” is equivalent to the assembly code call illustration of its usage is call 0040E204h, shown on line dword ptr ds: [a1a2a3a4a5a6a7a8] as shown in line 21. 27. This type of call statement can be either an API The equivalent API function name of the address function call or a procedure call. The control must be [a1a2a3a4a5a6a7a8], can be found in the displayed file returned back to the statement following where the call header of the portable executable program (see section was made. If this type of call statement leads to an API 4.1.2 and 4.1.3). The push statement displayed from function call, it must pass through a jmp statement, as lines 13, 15, 16, 17, 18, 19 and 20 are the arguments to shown on line 53. The memory address [0040E20Ch] of the API functions (seven arguments in this example). jmp statement on line 53, leads to the location were the There is a similar syntax of hexadecimal number “FF API function name can be found. The difference 15” on line 34, and the only difference is that the API between procedure call and jmp statement, is that the function address uses more than 7 push statement. When procedure call returns program control to the statement the statement “add esp, 4” is used, it means that there is following the call. The jmp statement never return a call 0040E204 in 27, inside another call dword ptr ds: control instead, it continues executing instruction [0040E20Ch] in line 34. The implication of having the following the target of the jmp location. statement add esp, 4 on line 28 immediately after the call on line 27, is because we are addressing a 32-bit

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

(iii) Target of Direct call of E8: The target of a procedure call, which has a hexadecimal number “E8” must use a ret (return) statement to pass the program control back to the statement after the call. The only way to know when to return a program control back to the statement following a procedure call statement, is to look for a return statement called ret. The function of the ret statement is to pass program control back to the statement following the procedure call. (iv) Indirect call of FF 15: The indirect call on lines 47 and 51, need a mov instruction on 43 and 45 to assign memory addresses to the call instruction. For instance, when the call esi instruction is encountered, the address is assigned to register esi and it needs to be extracted. When we move a bit up to line 43, we see the instruction mov esi, dword ptr ds: [01001030h]. This instruction means that the memory pointer address [01001030h] is assigned to the register esi, and this register is later called using the call statement on line 51.

4.1.2 Using Dumpbin to Display Portable Executable File Header 1 This section displays an illustration of how a file header of a portable executable program looks like. The illustrated executable file header and its set of API function names located in RAW DATA #1 section are displayed below. 31) a1a2a3a4a5a6a7a8 o1o2o3 API3 1) Dump of file filename.extension … 2) PE Signature found 32) name.dll can be any of msvcrt.dll, advapi32.dll, 3) File Type: EXECUTABLE IMAGE kernel32.dll, user32.dll, shell32.dll, 33) ntdll.dll. 4) FILE HEAD VALUES 34) 5) 14C machine (i386) 35) SECTION HEADER #2 6) case 5 3 number of sections 36) .data name 7) 3B7D849F time date stamp… 37) … 8) … 38) Read Write 9) OPTIONAL HEADER VALUES 39) RAW DATA #2 10) 10B magic # 40) a1a2a3a4a5000: h1h2 h3h4 h5h6 h7h8 h9h10 11) … h11h12 h13h14… h31h32 s1s2…s16 12) SECTION HEADER #1 41) a1a2a3a4a5010: h1h2 h3h4 h5h6 h7h8 h9h10 13) .text name h11h12 h13h14… h31h32 s1s2…s16 14) …. 42) a1a2a3a4a5020: h1h2 h3h4 h5h6 h7h8 h9h10 15) Execute Read h11h12 h13h14… h31h32 s1s2…s16 16) 43) … 17) RAW DATA #1 44) SECTION HEADER #3 45) .rsrc name 46) … 47) Read Only case 7 case 6 48) RAW DATA #3 18) a1a2a3a4a5000: h1h2 h3h4 h5h6 h7h8 h9h10 49) a1a2a3a4a5000: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 h11h12 h13h14… h31h32 s1s2…s16 19) a1a2a3a4a5010: h1h2 h3h4 h5h6 h7h8 h9h10 50) a1a2a3a4a5010: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 h11h12 h13h14… h31h32 s1s2…s16 20) a1a2a3a4a5020: h1h2 h3h4 h5h6 h7h8 h9h10 51) a1a2a3a4a5020: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 h11h12 h13h14… h31h32 s1s2…s16 21) … 52) … 53) Summary case 8 54) 2000 .data 22) section contains the following imports: 55) 2000 .rsrc 23) name.dll 56) 3000 .text 24) c1c2c3c4c5c6c7 Import Address Table 25) c1c2c3c4c5c6c7 Import Name Table

case 9 26) b1b2b3b4b5b6b7b8 time date stamp 27) b1b2b3b4b5b6b7b8 Index of first forwarder reference 28) case 10 case 11

29) a1a2a3a4a5a6a7a8 o1o2o3 API1 30) a1a2a3a4a5a6a7a8 o1o2o3 API2

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

(a) Machine Information: From line 1 to 11 contains a program data or code. The hexadecimal numbers are the information about the name of the program being placed in two pairs and a character space separating examined and the machine that is used to run the the pairs. The address takes record of the maximum program. The first part is the header information. The number of hexadecimal digits hk, where k is 1, 2, second part is the minimum processor type that was 3,…,32. used to run the program. The third part is the time the On the right side of hexadecimal numbers program was first executable on the computer system. are the set of symbols s1s2,s3,…,s16. These symbols The final part is the number of sections the header file display the numeric and alphabetic equivalent of the has and the time date stamp it takes. Line 1 shows the hexadecimal numbers hk. Line 23 shows the dll name filename and extension of the file in the form of whose API functions the executable program imports. filename.extension. Example of a file name including The code c1c2,c3,…,c7 of line 24 and 25 is the code its extension is edit.exe. PE (Portable Executable) in type of Import Address Table and Import Name Table. Line 2 shows that the program is a portable executable The code on line 25 and 26, which is 8-digits is file. associated with Index date stamp and Index of first b) Section Header: The section header has a section forwarder reference. The memory addresses ai on lines number, raw data numbers, raw data 29, 30, and 31 are equivalent contents of hexadecimal information and raw data contents. For each section, numbers hk of section 1 which represent addresses to the section and raw data numbers are the same. API function names. The ordinal numbers O1O2O3 Example of section numbers are shown in lines 12, 35, are hexadecimal numbers which act as beginning and and 44. Example of raw data numbers are shown in end markers for to identify an API function equivalent lines 15, 39, and 48. In the raw data contents, names from the hexadecimal numbers hk of section 1 a1a2a3a4a5000: is an 8-digit address of 32-bit of Dumpbin paragraph. hexadecimal data and a semicolon. The ai is a iii) Summary Information: The section list all dexadecimal number where i can be 1, 2, 3, 4 or 5. components of executable program whose The last digit of the address is a hexadecimal contents is being displayed. Examples of 0. After the semicolon, is a set of hexadecimal the list of file header summary are .data and .text numbers h1h2 h3h4 h5h6…h31h32 that can represent

4.3 Using Dumpbin to Display Portable Executable File Header 2 This section shows the contents of a file header of an illustrated portable executable program whose set of API function names are located in RAW DATA #3. The displayed file header contents are shown below.

1) Dump of file filename.extension 2) PE Signature found 3) File Type: EXECUTABLE IMAGE 4) FILE HEAD VALUES 5) 14C machine (i386) 6) 4 number of sections 7) 0 time date stamp… 8) … case 12 9) OPTIONAL HEADER VALUES 10) 10B magic # 11) … 12) SECTION HEADER #1 13) .text name 14) …. 15) Execute Read case 13 case 14 16) 17) RAW DATA #1 18) a1a2a3a4a5000: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 s1s2…s16 19) a1a2a3a4a5010: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 s1s2…s16 20) a1a2a3a4a5020: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 s1s2…s16 21) … 22) RAW DATA #2

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

case 15 case 16

23) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 … h31h32 API1 s1A 24) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 PI2s2s3 25) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 s4API3 26) … 27) RAW DATA #3 case 18 case 17

28) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 API1 s1A 29) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 PI2s2s3 30) a1a2a3a4a5a6a70: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12… h31h32 s4API3 31) … 32) Dump of file filename.extension 33) section contains the following imports: 34) name.dll 35) c1c2c3c4c5c6c7 Import Address Table 36) c1c2c3c4c5c6c7 Import Name Table 37) 0 time date stamp 38) case 19 0 Index of first forwarder reference 39) 40) O1O2O3 API1 41) O1O2O3 API2 42) O1O2O3 API3 43) … 44) name.dll can be any of msvcrt.dll, gdi32.dll, comctl32.dll, advapi32.dll, kernel32.dll, 45) user32.dll, shell32.dll, ntdll.dll. 46) 47) … 48) RAW DATA #4 49) a1a2a3a4a5000: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 50) a1a2a3a4a5010: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 51) a1a2a3a4a5020: h1h2 h3h4 h5h6 h7h8 h9h10 h11h12 h13h14… h31h32 s1s2…s16 52) 53) Summary case 20 54) 2000 .bss 55) 3000 .data 56) 1000 .idata 57) 1000 .rdata 58) 1000 .reloc 59) 1000 .rsrc 70) 7000 .text

The file header 2 is similar to file header 1. The differences are: (i) in the machine information part, time data stamp stores the value 0 in file header 2, (ii) there are more than 3 section headers in file header 1, (iii) the Dynamic Link Library (DLL) name is located in section and RAW DATA #3 of file header 2, (iv) the code associated with index date stamp and index of first forwarder reference is 0 in file header 2, (v) the API names (line 40, 41, and 42) are associated with only ordinal numbers and no addresses in file header 2, (vi) the API functions can be identified on the hexadecimal numbers hk and the alphabetic equivalent is displayed in the symbols section of the RAW DATA #3 in file header 2.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

4.4 The Modification of Dumpbin Microsoft Disassembly Case 10: The label case 10 on line 29 of section 4.1.2 Portion uses the memory address to read in the API function The modification of Microsoft Dumpbin Disassembly portion name sent from case 7. is an attempt to find and replace the API function pointer address, with its equivalent function name. Case 11: The label case 11 on line 29 of section 4.1.2 gives the API function name identified by its address. 4.4.1 The Features of a Disassembled Program Contents The features of a disassemble program contents are explained Case 12: The label case 12 on line 7 of section 4.1.3 by attaching a case label to the different section of the checks if a character value ‘0’ is assigned to the string disassembled program. The disassembled contents are the “time date stamp”. results of disassembly of a portable executable program using a Microsoft Dumpbin disassembly tool. This is explained Case 13: The label case 13 on line 18 of section 4.1.3 using each case label attached to the disassembled executable contains an equivalent of a API function address in code as follows: square bracket. The last digit of the address acts as the character distance position pointer, where 4 bits of Case 1: The label case 1 on line 21 and 34 of section hexadecimal numbers would be read in reverse 4.1.1 shows the call instruction type of “FF 15” with its direction. memory address in square brackets. Case 14: The label case 14 on line 18 of section 4.1.3 is Case 2: The label case 2 on line 43 and 45 of section similar to the description on case 13. The only 4.1.1 show the movement of the memory address in difference is that 5 bits of hexadecimal numbers would square brackets into registers esi and edi. be read in a reversed direction instead of 4 bits.

Case 3: The label case 3 on line 27 of section 4.1.1 Case 15: The label case 15 on line 23 of section 4.1.3 shows the call instruction type of “E8” with its memory attempts to read the last 4 bits of address. This address address. leads to an API function name of the address read in reversed direction of case 13 on line 17. Case 4: The label case 4 on line 53 of section 4.1.1 shows the jmp instruction type of “FF 25” with its Case 16: The label case 16 on line 23 of section 4.1.3 memory address in square. reads the distance character position pointer using the last address digit of the address of case 15 on line 23. Case 5: The label case 5 on line 7 of section 4.1.2 shows the string “time date stamp” and its address. Case 17: The label case 17 on line 28 of section 4.1.3 is similar to the description of case 15 of line 23. The only Case 6: The label case 6 on line 18 of section 4.1.2 difference is that it reads the last 5 bits of the memory shows section 4.2 the equivalent memory address by address. displaying the last digit as a ‘0’. Case 7: The label case 7 on line 18 of section 4.1.2 Case 18: The label case 18 on line 28 of section 4.1.3 is attempts to trace the 8 digits address of case 6, including similar to the description of case 16 of line 23. The only the last digit number from the hexadecimal number difference is that the last digit of the memory address is section in a reversed form. used as the character distance position pointer, for address of case 17. Case 8: The label case 8 on line 22 of section 4.1.2 attempts to identify the string “section contains” so that Case 19: The label case 19 on line 19 of section 4.1.3 to determine the area where the set of API function checks the string “time date stamp” for it character value names are located. ‘0’. Case 20: The label case 20 on line 53 of section 4.1.3 Case 9: The label case 9 on line 26 of section 4.1.2 shows the summary of the contents of each section. checks if a character value ‘0’ is the address of the string “time date stamp”.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

4.5 Summary of Procedure for Modifying Microsoft Dumpbin This kind of program construct cannot be revealed by Disassembler conventional disassembly software tools and malicious program In determining API call from File header 1, in section 4.1.2, when writers can use this technique to make program understanding an API function address is captured in case 1, the first digit is very difficult. Also, antivirus experts can now conveniently use checked if it is 0. When it is a 0 of case 8 and the last value of the Microsoft Dumpbin software tool as a plug-in into their antivirus address of case 5 is not 0, then the character position pointer system because, the memory pointer addresses has been replaced moves to case 6. In case 6 the address captured is traced ignoring with API function names. the last digit. When the address is seen, its last digit is used to move the character pointer to the position where it will read a 6. CONCLUSION memory address that will lead to the API function name. The position moved to by the character pointer reads the 8 bits address In this paper, an attempt is made to modify Microsoft Dumpbin in a reverse direction of case 7. The address captured by reading a disassembly software tool. The tool also has the ability to expose hexadecimal number in case 7, leads the character pointer to the detailed nested procedural calls, which other contemporal position where the address of the API function name can be found. disassembly software tools cannot reveal. Dumpbin has two major disadvantages, the first one is that it cannot display API function For File header 2 in section 4.1.3, when an API function address is name instead, it shows its equivalent memory pointer address. The captured in case 1, the first digit is checked if it is 1. When it is 1 second problem is that it is built to be used as a console programs and the last value of the address of case 12 is 0, then the character analysis tool and not as a programming language plug-in. position pointer moves to case 13. In case 13 the address captured is traced ignoring the last digit. When the address is seen, its last This research solves the first problem and proposes how the digit is used to move the character pointer to the position where it second problem can be solved. To solve the first problem, will read a memory address that will lead to the API function Dumpbin software tool was modified to identify API function name. The position moved to by the character pointer, reads the 4 name equivalent of a memory pointer address. To solve the second bits address in a reversed direction of case 15. The address problem, a procedure was given to describe how to identify a captured by reading a hexadecimal number in case 15, leads the memory pointer address and replace it with its equivalent API character pointer to the hexadecimal number section of address in function name. This paper also gave a technique of how to resolve case 16. In the hexadecimal number section, attempt is made to complex nested procedural calls in a disassembled program read the hexadecimal number equivalent of the English alphabet to contents. We hope that malicious programs analysts will use the get the API function name. procedure of modified Dumpbin software to enhance the performance of their malicious program analysis. Also, for File header 2, when an API function address is captured in case 1, the first digit is checked if it is neither 0 nor 1. When it is neither 0 nor 1, it has the similar principle as when the digit is 1. REFERENCES The only difference is that the position moved to by the character [1] Eagle C. (2011): The IDA Pro Book, Second Edition, No pointer reads the 5 bits address in a reversed direction of case 15. Starch Press, Inc., Francisco, U.S.A. The address captured by reading a hexadecimal number in case 15 lead the character pointer to another hexadecimal number section [2] Edmond J. M (2007): Counterintelligence through of address in case 17. In the next hexadecimal number section in Malicious code analysis, Master Thesis, Naval Post case 18, attempt is made to read the hexadecimal number Graduate school, Monterey. equivalent of the English alphabet to get the API function name. [3] Eilam E. (2005): Reversing: Secrets of Reverse 5. DISCUSSION Engineering, Wiley Publishing, Inc., Indianapolis, Indiana, pages 109-138. The procedure that is used to modify Microsoft Dumpbin disassembly tool so that memory pointer addresses are replaced [4] Hyde R. (2006): Write Great Code, Volume 2: Thinking with its equivalent API function name, is a significant effort. Low-Level, Writing High-Level, No Starch Press, Inc., Microsoft Dumpbin software tool is used as a malicious program Francisco, U.S.A. analysis tool at console level. That is, the software tool was not built to be used as a plug-in to a programming language written to [5] Kaspersky K. (2003): Hacker Disassembling Uncovered, A- perform malicious program analysis /detection. The modification LIST Publishing. of Microsoft Dumpbin software tool can make malicious programs analysts to use the tool with its API functions visible. Another [6] Kinder J. (2005): Model Checking Malicious Code, advantage of using a modified Microsoft Dumpbin software tool, Fakultat Fur Informatik, Technische Universitat is that reverse engineers can now view constructs of system Munchen, Germany. function called inside another function. [7] Konstantin R. (2005): Efficient Static Analysis of Executables for DetectingMalicious Behaviours, M.Sc. Thesis, Polytechnic University, Brooklyn, NY, U.S.A.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[ 8] Mohammad N. S. (2009): Malware Analysis, B.Sc Thesis, Helsinki Metropolia University of Applied Sciences. Author’s Brief

[ 9] Mohamad F. Z and Aman J(2011). An Approach for Osaghae Osakpanwan Edgar obtained B.Sc Identifying Malware Operation and Target using Run Time and M.Sc Degrees in the Department of Analysis and Resource Monitoring, International Journal of Computer Science, University of Benin, Digital Content Technology and its Applications, Vol. 5, Nigeria. He is also a Doctoral student in the Number 8, pages 169-178. Department of Computer Science of the same University. He is presently a Faculty member [10] Murugen S and Kuppusamy (2011): Malware Analysis in the Department of Computer Science, Using Assembly Level Program, International Journal of University of Port Harcourt, Nigeria. His Advanced Engineering Sciences and Technologies, Vol. 2, research has focused on Computer Virology, Number 1, pages 1-12. Formal Verification methods and Computer Algorithms. He can be reached by phone on +2348034931540 and through E-mail [11] Sami A., Yedegari B., Rahimi H., Peiravian N., Hashemi S. [email protected]. and Hamze A. (2010): Malware Detection Based on Mining API Calls, Proceedings of the 2010 ACM Symposium on Applied Computing (SAC), Sierre, Switerland, Pages 1020- 1025.

[12] Willems C., Holz T and Freiling (2007): Toward Automated Dynamic Malware Analysis Using CWSandbox, IEEE Security and Privacy,Vol. 5, Number 2, Pages 32-39.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Verifying and Validating Recursive Performance Models of Parallel Computer System Using Z-Transform

1O.E. Oguike, 2S.C.Echezona & 3M.N. Agu Department of Computer Science University of Nigeria Nsukka, Nigeria [email protected], [email protected], [email protected]

ABSTRACT Recursive models have been used in the past to model the variation of number of processes in all the queuing systems of heterogeneous parallel computer system. The main reasons for the use of recursive models when modeling the variation of number of processes in all the queuing systems of a heterogeneous parallel computer system is the inability of the analytic technique to determine the exact convergence of some mathematical series. The quality of this novel technique of modeling the performance of parallel computer system needs to be assessed. The assessment of the quality of recursive approach to performance modeling can be done by verifying and validating the model, which will help to assure us of the accuracy and correctness of the model. However, an alternative to the use of the recursive models when modeling variation of a performance metric is the statistical technique of the Z-Transform. The Z-Transform was developed as a result of the inability of the analytical technique to determine the exact convergence of some mathematical series. This paper uses the Z-transform to validate or assess the quality of the recursive approach to modeling variation of a performance metric of parallel computer system.

Keywords- Recursive models; performance modeling; Z-Transform; parallel computer; distributed memory; parallel computer system; queuing network; variations; model verification; model validation.

1. INTRODUCTION We assume that processes arrive at the various queues according to Poisson distribution, and they are serviced according to Model verification and validation can be used to assess the quality Exponential distribution [5, 6]. of simulation models. Model verification and model validation are essential parts of model development, which help to assess the 2. LITERATURE SURVEY quality of the developed models. If a model is not verified and validated it cannot be assured of quality, therefore, it can be sent The author in [36] pointed out that no computational model will back to the drawing board. Model verification is done in order to ever be fully verified. He pointed out that a high degree of ensure that the simulation algorithm used to simulate the model on statistical certainty could be realized for a model as more cases of the computer is correct and the simulation program is correctly the model are being tested. The authors in [33] pointed out that programmed. Model verification eliminates every error that may one of the ways to carry out the verification of a model is through occur when implementing the models on the computer. On the extensive testing. According to them, this is necessary because it other hand, model validation aims at making the model address the will ensure that all conditions that could arise in simulation right problem, address accurate information about the system operation have been covered by test cases. Furthermore, according being modeled. Model validation compares the results of the to [36], the reason for developing some models is for insight, not simulated models with the results of real simulated system. numbers, such models will help to gain insight into key variables and their causes and effects. Therefore, model validation tries to establish if the simulation model is an accurate representation of the real system. However, The authors in [33] argued that though quantitative comparison for some reasons, it may not be easy sometime to obtain results of will provide the basis for validation, however, it can miss the the real system, in such a situation, expert knowledge can be used qualitative discrepancies or agreements that human are capable of to determine if the qualitative data from the simulated model detecting. They suggested two methods that can be used for represents the real system or if it does not represent the real detecting such discrepancies or agreements, which are: system [34]. Model validation and verification will be applied to visualization and animation. Visualization, according to them the models developed in [35] for heterogeneous parallel computer helps to map numerical data into graphical structure that human system with distributed memory, as shown in figure 1. We assume can more readily understand. that the various queues are finite [1, 2, 3, 4] i.e. there is a limit to the number of jobs that can be admitted into the queues, and This graphical display of the results of the simulated model or the negligible communication overhead. Suppose X1, X2, X3, … , Xn , system behavior will help us to determine if the model is valid or Xn+1, Xn+2, Xn+3, …, Xn+K are the maximum number of processes invalid. Furthermore, [33] pointed out that quantitative that can be admitted into the respective queues. comparison is needed to make finer distinctions between behaviors that agree in their basic form, but qualitative comparison can help to eliminate models that are not in the right

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

ballpark. Sometimes a validated model can be used to validate Sum1= Term1(X, ) + Sum1(X-1, ) another model by comparing qualitative and quantitative data of 3. Display Sum1 the two models [36]. Term1(X, ) is the recursive model that determines the xth term 3. VALIDATING RECURSIVE MODELS USING Z- of a series, it is given as: TRANSFORM

1, Xi = 0 Suppose Xi denotes the maximum number of processes that can be in the ith finite queuing system at any time [12, 13, 14], and  i denotes the utilization factor for the ith queue. When the queuing Term1i(Xi,  i ) = (3) network is considered, the utilization factor for the ith queuing system is modeled in [35] as: i * Term1i(Xi-1,  i ), Xi 0    , i  0 The following recursive algorithm can be used to implement the  0 i recursive model in equation (3). i     i , i  1,2,3,..., n  k Double Term1(int X, double )    0 i 1. Request data (1) 1.1 Request X 1.2 Request  is the arrival rate from the outside world into the queuing 2. Determine Term1 network, and is the probability that a process will be scheduled  i 2.1 Term1 = 1 if X = 0 else into the ith queuing system and is the probability that a process  0 Term1 = Term1i(Xi-1, ) will depart from the queuing network after each cpu burst, while 3. Display Term1

 i is the departure rate for the ith processor, n denotes the number of parallel processors, while k denotes the number of optional I/O The recursive model, called Sum2i(Xi,  i ) is given as: processors. 1, Xi = 1

The following recursive models have been developed in [35], which were used to model the variation of a performance metric of (4) heterogeneous parallel computer system with distributed memory:

1 , X = 0 Term2 (X )*term1 (X -1, ) + Sum2 (X -1, ), X 1 i i i i  i i i  i i

The following recursive algorithm can be used to implement the Sum1(X,  ) = (2) recursive model in equation (4).

Double sum2(int X, double ) Term1(X,  ) + Sum1(X-1, ), X  0 1. Request data 1.1 Request X The following recursive algorithm can be used to implement the 1.2 Request recursive model in equation (2). 2. Determine Sum2 2.1 Sum2 = 1 if X = 1 else

Sum2= Term2i(Xi)*term1i(Xi-1, ) + Sum2i(Xi-1, )

3. Display Sum2 Double sum1(int X, double  ) The recursive model, Term2i(Xi) is given as:

1. Request data 1.1 Request X 1.2 Request  2. Determine Sum1 2.1 Sum1 = 1 if X = 0 else

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

1, Xi = 1 The recursive model, Term3i(Xi) has been defined in [35] as:

1, Xi = 1 Term2 (X ) = i i (5) Term3i(Xi) = (8)

1 + term2i(Xi-1), Xi  1

(2*Xi-1) + Term3i(Xi-1), Xi 1 The following recursive algorithm can be used to implement the recursive model in equation (5). The following recursive algorithm can be used to implement the recursive model in equation (8). Double Term2(int X) 1. Request X Double Term3(int X) 2. Determine Term2 1. Request X 2.1 Term2 = 1 if X = 1 else 2. Determine Term3 Term2 = 1 + term2i(Xi-1) 2.1 Term3 = 1 if X = 1 else 3. Display Term2 Term3 = (2*Xi-1) + Term3i(Xi-1) The recursive model, term1 (X , ) is defined in equation (3). i i  i 3. Display Term3 Therefore, combining equations (2) and (4), this recursive model has been obtained in [35]. Therefore, combining equations (7) and (2), this recursive model given in equation (9) has been developed in [35].

 Sum2i (X i , i )  E(xi )    (6)   Sum3(X ,  )   Sum1 (X ,  )  2  i i i   i i i  E(xi )    (9)  Sum1(X i , i )  Therefore, combining equations (6) and (9), the recursive model X is a random variable that denotes the number of processes in i given in equation (10) has been developed in [35]. the ith queuing system. The following recursive model called 2  iSum3(X i , i )   iSum2(X i , i )  (10) VAR(X )       Sum3i(Xi,  i ) has been developed in [35], and it is given as:      Sum1i (X i , i )   Sum1i (X i , i )  Furthermore, the recursive model given in equation (11) has been 1, Xi = 1 developed in [35]. 2 1  nk   Sum3(X , )    Sum2(X , )   VAR(Y)    i i i    i i i  (11) (n  k)2  Sum1 (X , )   Sum1 (X , )   (7)  i1  i i i   i i i   Y is the random variable that denotes the average number of processes in all the queuing systems of the queuing network, and equation (11) models the variation of average number of processes Term1 (X -1,  )*Term3 (X ) + Sum3 (X -1,  ) ,X  1 i i i i i i i i i in all the queuing systems of the queuing network. The Z transform will be used to validate the model in equation The following recursive algorithm can be used to implement the (11). The Z-transform is defined inx [15] as: recursive model in equation (7) is given as: GXi(z) = E(z ) (12) Double sum3(int X, double  ) 1. Request data 2  G Xi (z) 1.1 Request X VAR( X i) = z=1 + 1.2 Request  (z) 2 2. Determine Sum3 2.1 Sum3 = 1 if X = 1 else  GXi (z)    z=1 - Sum3 = Term1i(Xi-1, )*Term3i(Xi) + Sum3i(Xi-1, )  (z)  3. Display Sum3 2  G Xi (z)    z=1 (13)  (z) 

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The expectation has been defined in [15, 23] , statistically as: G (z)   1 X X 1   X (X 1)  Xi     (21) X i 1   X i z   1 i  (1 )  E( xi ) = xi Pi . (14)  xi xi 0 Furthermore, taking the second derivative of equation (19), with respect to z and initializing z to 1, we obtain the following: The probability density function for the ith queuing system with Xi maximum number of processes in the queuing system and utilization factor for the ith queue is , has been defined in [7] 2   GXi (z)  (1 i )  u  v      (22) as: 2  X i 1  4  (z) 1 i  (1 )   xi (1  ) i i Simplifying further, u and v are given as:  , i 1, x  0,1,2,3,...X i X i 1  1 i P  2 X i 2 2 X i 1 X i 1 x i  u  1   X (X 1)  X   X  (23) 1 i  i i i i i i i   , i 1  (15) X 2 X 1 X 1  X i 1 i i i v  2i 1 i Xi  X i i  i  i  (24)

Simplifying equation (13) further, using equations (14) and (15), Therefore, using equations (24) and (23) in equation (22), and for i 1, we obtain the following: using equations (22) and (21) in equation (13), we obtain the Z- Transform model for the variation of number of processes in the ith queuing system. Furthermore, the Z-Transform can be used to X x i   i (1  )  obtain variation of the average number of processes in all the xi  i i  GXi (z)  z (16)   X i 1  queuing systems of the queuing network, as shown in equation (25) xi 0  1 i  below.

Simplifying further, we obtain the following: 1  nk  VAR(Y)   VAR(X ) (25) 2  i X i (n  k)  i1   (1  )  x  i  xi i GXi (z)  z i  X i 1  1  x 0  i  i (17) However, for i  1, we obtain the expectation and variance analytically without the use of recursive models, therefore, there Simplifying further, we obtain the following: may not be any need for the Z-Transform approach. In order to compare the qualitative and quantitative results of the simulation recursive models and the Z-Transform models, the two sets of  (1  )  X i  i  xi models have been simulated and the results of the simulation are GXi (z)  (zi )  X i 1  presented below. 1  x 0  i  i (18) 4. SIMULATION RESULTS OF THE RECURSIVE Simplifying further, we obtain the following: MODELS AND Z-TRANSFORM MODEL X 1 The recursive models and Z-transform model were converted to  (1  )  1 (z ) i   i   i  computer programs, using Java programming language and the GXi (z)    1  X i 1  (1 (z ))  results of simulation were analyzed to determine how performance  i  i  (19) metric under consideration changes as a particular parameter varies, while other parameters remain constant [10]. Table 1 shows Therefore, taking the first derivative of equation (19), with respect the result of the simulation, using a two-processor parallel to z, and initializing z to 1, we obtain the following: computer system. If the probability of a process leaving the system is known to be 0.2 and the probability that a process will join the G (z)  (1  ) 1 X X 1   X (X 1)  first and second queues are 0.775 and 0.025, respectively. Suppose Xi   i   (20) the first processor is a high-speed processor with high departure  X i 1  2  z 1   (1 )  rate of 30, while the second processor is a low speed processor with  i  a low departure rate of 10. Suppose maximum number of processes to be allowed into first queue is 20, while maximum number of Simplifying further, equation (20) reduces to: processes to be allowed into the second queue is 5. The experimental trials were carried out several times, in each trial, the

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

arrival rate was changed, and the corresponding variation was departure rate increaes, the the variation decreases accordingly, and obtained as the result of the simulation. The result can be presented as the departure rate decreases, the variation increases accordingly. in table 1 and in figure 2. 5. SUMMARY AND CONCLUSION The result shows that the behaviour of the variation for non- compute intensive applications (i.e. when the utilization factor for This paper has been able to validate the developed model for the the ith queuing system is less than 1) is different from its variation of performance metric of heterogeneous parallel computer behaviour for compute intensive applications (i.e. when the system with distributed memory by using Z-transform. The models utilization factor for the ith queuing system is greater than 1), for have been simulated on the computer and the results of the the various values of the arrival rates. The results show that for simulat1on for the developed models and the Z-transform show non-compute intensive applications, increasing the arrival rate will that both the quantitative and qualitative results of the Z-transform lead to a corresponding increase in the variation of the models are the same as that of the developed models. Therefore, performance metric of the heterogeneous parallel computer, while with the result, it has been discovered when the heterogeneous decreasing the arrival rate will lead to a corresponding decrease in parallel computer system with distributed memory will realize the variation of the performance metric of heterogeneous parallel minimum variation. computer system. On the other hand, for compute intensive applications, increasing the arrival rate will lead to a REFERENCES corresponding decrease in the variation, while decreasing the arrival rate will lead to a corresponding increase in the variation. [1] Henry H. Liu and Pat V. Crain, An Analytic Model for Predicting the Performance of SOA-Based Enterprise Furthermore, as we keep the following input parameters constant Software Applications, Proc. International Conference for non-compute intensive applications, the probability that a of Computer Measurement Group, (2004). process will leave the network is 0.2, the probability that a process [2] S. Balsamo et al, A Review of Queueing Network will join queue 1 and 2 are 0.775 and 0.025, respectively. The Models with Finite Capacity Queues for Software departure rate for processor 1 and 2 are 30 and 10, respectively, Architecture Performance Prediction, (2002). and the arrival rate from the outside world is 4. By varying the [3] Catalina M. Liado et al, A Performance Model Web degree of multiprogramming (maximum number of processes in Service, Proc. International Conference of Computer the system) for the two queues of a two-processor parallel Measurement Group, (2005). computer system, we obtain the corresponding variation shown in [4] Rosselio, J et al, A Web Service for Solving Queueing table 2 and figure 3. Network Models Using PMIF. www.perfeng.com/paperndx.htm, (2005). Similarly, as we keep the following input parameters constant for [5] Cathy H. Xia, Zhen Liu., Queueing systems with long- compute intensive applications, the probability that a process will range dependent input process and subexponential leave the network is 0.2, the probability that a process will join service time. Proc. ACM SIGMETRICS international queue 1 and 2 are 0.775 and 0.025, respectively. The departure rate conference on Measurement and modeling of computer for processor 1 and 2 are 30 and 10, respectively, and the arrival systems,(2003). rate from the outside world is 400. By varying the degree of [6] Shanti Subramanyam, Performance Modelling of a J2EE multiprogramming (maximum number of processes in the system) Application to meet Service Level s, Agreement, Proc. for the two queues of a two-processor parallel computer system, we International Conference of Computer Measurement obtain the corresponding variation shown in table 3, and figure 4. Group, (2005) The interpretation of the results in figure 3 and 4 is that for both [7] Hamdy A. T.,. Operation Research: An Introduction, compute and non-compute intensive applications, the behavior of Prentice-Hall of India, (1999). the variation is the same as the total maximum degree of [9] Ivan Stojmenovic; Recursive Algorithms in Computer multiprogramming changes. In a similar manner, as we keep the Science Courses : Fibonacci Numbers and Binomial following input parameters constant, probability of a process Coefficients; IEEE Transactions on Education; Vol. 48, leaving the network is 0.2, while the probability of a process going No. 3 to queue 1 and 2 is 0.4. The maximum number of processes that [10] Arjan J.C. van Gemund; Performance Modelling of can be in queue 1 and 2 are 15 and 14, respectively., and the arrival Parallel Systems: An Introduction. rate from the outside world is 4. By changing the departure rates of [11] Justyna Berlinska, The Statistical models of parallel the two processors, we obtain the corresponding variation of the applications, Annales UMCS Informatica, (2005). performance metric, as shown in table 4 below. [12] Arranchenkov, K.E., Vilchersky, N.O., Shevlyakor, G.L The result shows that the behaviour of variation for compute Priority queueing with finite buffer size and randomized intensive applications is different from the behaviour of variation push-out; mechanism. Proc. of ACM SIGMETRICS for non-compute intensive applications, for different values of the international conference on measurement and modeling total departure rates. From the results in table 4 and figure 5, we of computer systems.; (2003). discover that for compute intensive applications, as the total departure rate increases the variation increases as well, and as the departure rates decreases, the variation decreases as well. On the other hand, for non-compute intensive applications, as the

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[13] Abunday, B.D., and Khorram, E. The finite source [35] O.E. Oguike et al; Modeling Variation of Performance queueing model for multiprogrammed computer systems Metric of Distributed Memory Heterogeneous Parallel with different CPU times and different I/O times. Acta Computer System Using Recursive Models; 3rd IEEE Cybern. 8, 4 , (1998) International Conference on Computational Intelligence [14] J. Sztrik; Finite-Source Queueing Systems and their Modeling and Simulation; (2011). Applications: A Biliography; [36] Charles M. Macal; Model Verification and Validation; [15] Trivedi K. Shridharbhai, Probability and Statistics with Workshop on “Threat Anticipation: Social Science Reliability, Queuing and Computer Science Methods and Models”; The University of Chicago and Applications, John Wiley & Sons Inc., (2002). Argonne National Laboratory; (2005). [16] Per Brinch Hansen. Operating System Principles. Prentice-Hall of India Private Limited, (1990). [20] J. Sztrika and T. Gál A recursive solution of a queueing model for a multi-terminal system subject to breakdowns; Performance Evaluation Volume 11, Issue cpu queue 1, Published by Elsevier, (1990). Parallel [23] Robert V. Hogg and Allen T. Craig; Introduction to Pocessors Mathematical Statistics; Macmillan Publishing Co. Inc.; (1978). cpu queue [24] Andrea Clemantis, Angelo Corana; Modelling Performance of Heterogeneous Parallel Computer System; Journal of Parallel Computing, Volume 12, cpu queue Issue 9, Elsevier; pages 1131-1145; (1999). [25] E. Post, H.E. Goosen; Evaluating the Parallel Performance of a Heterogeneous System cpu queue [26] Beutler, F; Mean sojourn times in markov queuing network: Little’s formula revisited; IEEE Transaction on

Information Theory; Volume 29, Issue 2, page 233-241; (2003). I/O queue [27] Ken Vastola; http://networks.ecse.rpi.edu/~vastola/pslinks/perf/node4 6.html I/O queue [28] Xiaodong Zhang, Yong Yan; Modeling and Characterizing Parallel Computing Performance on

Heterogeneous Network of workstations; Proceedings of I/O queue the 7th IEEE Symposium on Parallel and Distributeed Processing (SPDP ’95) 1063-6374/95 $10.00 © 1995 IEEE I/O processors [29] O.E. Oguike et al; Modelling the Performance of Computer Intensive Applications of Parallel Computer nd Figure 1: Queuing network of a heterogeneous parallel computer system System; Proc. Of IEEE 2 International Conference on with distributed memory. Computational Intelligence, Modeling and Simulation; (2010). [30] O.E. Oguike et al; Evaluating the Performance of

Parallel Computer System Using Recursive Models; Proc. Of IEEE 4th UKSim European Modeling Symposium; (2010). [31] O.E. Oguike et al; Evaluating the Performance of Heterogeneous Distributed Memory Parallel Computer System Using Recursive Models; 2nd IEEE International Conference on Intelligent Systems, Modeling and Simulation; (2011). [32] Leonard Kleinrock, Queueing Systems Volume 1 and 2, John Wiley & Sons, (1975). [33] Bernard P. Zeigler et al; Theory of Modelling and Simulation; Elsevier; (2000) [34] Cor van Dijkum et al; Validation of Simulated Models; Siswo Publication 403, Amsterdam, (1999)

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

TABLE 1: RESULT OF VARIATION AGAINST ARRIVAL RATE Key to the table: AR VFM VFZT MPQ1: Maximum Number of Processes in Queue 1 1 0.05 0.05 MPQ2: Maximum Number of Processes in Queue 2 2 0.12 0.12 VFM: Variation from Model 3 0.27 0.27 4 0.57 0.57 VFZT: Variation from Z-Transform 5 1.29 1.29 Variation Against Maximum Number of 6 3.32 3.32 Processes 7 7.45 7.45 8 8.98 8.98 0.6 9 5.95 5.95 0.5 10 3.32 3.32 0.4 0.3

11 1.98 1.98 0.2 Variation 12 1.32 1.32 0.1 13 0.97 0.97 0 0 10 20 30 40 50 14 0.76 0.76 Maximum Number of Processes 15 0.62 0.62 Key to the table: Figure 3: Validated Variation Against Maximum Number of AR: Arrival Rate Processes for Non-Compute Intensive Applications

VFM: Variation from Model TABLE 3: RESULT OF VARIATION AGAINST MAXIMUM NUMBER OF PROCESSES FOR COMPUTE INTENSIVE APPLICATIONS VFZT: Variation from Z-Transform MPQ1 MPQ2 VFM VFZT Variation Against Arrival Rate 3 5 0.08258 0.08258 4 6 0.083 0.083 10 9 5 8 0.08315 0.08315 8 8 11 0.08316 0.08316 7 6 11 14 0.08316 0.08316 5 4 14 17 0.08316 0.08316

3 Variation 2 17 20 0.08316 0.08316 1 0 20 23 0.08316 0.08316 0 2 4 6 8 10 12 14 16 Key to the table: Arrival Rate MPQ1: Maximum Number of Processes in Queue 1 MPQ2: Maximum Number of Processes in Queue 2 Figure 2: Validated Variation Against Arrival Rate. VFM: Variation from Model TABLE 2: RESULT OF VARIATION AGAINST MAXIMUM NUMBER OF PROCESSES FOR NON-COMPUTE INTENSIVE APPLICATIONS VFZT: Variation from Z-Transform MPQ1 MPQ2 VFM VFZT 3 5 0.236 0.236 5 8 0.389 0.389 8 11 0.51 0.51 11 14 0.554 0.554 14 17 0.564 0.564 17 20 0.566 0.566 20 23 0.567 0.567

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Variation Against Maximum Number of Variation Against Total Departure Rate Processes for Compute Intensive Applications

9 0.0832 8 0.0831 7 0.083 6 0.0829 5

0.0828 4

Variation 0.0827 3 Variation 0.0826 2

0.0825 1 0 10 20 30 40 50 0 Maximum Number of Processes 0 10 20 30 40 50 60 Departure Rate

Figure 4: Validated Variation Against Maximum Number of Processes for Compute Intensive Applications. Figure 5: Validated Variation Against Departure Rate.

TABLE 4: RESULT OF VARIATION AGAINST DEPARTURE RATE

DRP1 DRP2 VFM VFZT Authors’ Brief 1 3 0.28 0.28 Oguike, Osondu Everestus is a Senior 3 5 1.3 1.3 Lecturer in the Department of Computer 5 7 4.93 4.93 Science, University of Nigeria, Nsukka, 7 9 8.3 8.3 Enugu State, Nigeria. He has received many academic prizes and scholarships as a result 9 11 6.45 6.45 of his outstanding academic performance. He 11 13 3.05 3.05 is interested in modeling the performance of 13 15 1.62 1.62 parallel computer system. He can be reached by phone on +2348035405100 and through E-mail [email protected] 15 17 1.03 1.03 17 19 0.73 0.73 19 21 0.56 0.56 Dr (Mrs) Monica N. Agu is of Department of 21 23 0.45 0.45 Computer Science, University of Nigeria, Nsukka, in the faculty of Physical Sciences. 23 25 0.38 0.38 Her research has focused on using Information 25 27 0.32 0.32 and Communication Technology on Poverty Key to the table: Alleviation and Modelling the performance of Computer Systems. She can be reached by DRP1: Departure Rate for Processor 1 phone on +2348039329480 and through E-mail DRP2: Departure Rate for Processor 2 [email protected]

VFM: Variation from Model VFZT: Variation from Z-Transform Echezona Stephenson C. is a faculty member, as well as, on a PhD of the Computer Science Department, University of Nigeria, Nsukka. His research has focused on using Z-Transform to corroborate the Recursive Performance Models of Parallel Computer Systems. He can be reached by phone on +23437325573 and e-mail through [email protected] and [email protected].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Improved Electoral Fraud Prevention Mechanism for Efficient and Credible Elections in Nigeria

1F.O. Aranuwa & 2O. Oriola Department of Computer Science, Adekunle Ajasin University Akungba – Akoko, Nigeria [email protected], [email protected]

ABSTRACT We defined voting system as a vital tool to collect and reflect people’s opinion to elect a candidate of their choice to occupy a position or people to govern them. Obviously, its importance in a democratic system or society cannot be over-emphasized. However, voting system in developing countries, especially in Sub-Sahara Africa, Nigeria inclusive has witness lot of challenges and electoral fraud during their elections. Commonly acknowledged types of vote fraud and challenges are: voter registration fraud, dead people voting, voter suppression, voter caging and purging, voter impersonation, voting more than once, election rigging, result prevarication and litigation after elections. All these, apart from the economic implications will not only lead to irregular electoral process and instability, but also make the electorates to lose confidence in the electoral system. Hence, the need for an improved automated electoral mechanism. In this work, we propose an Automated Direct Recording Electronic Voting System (ADREVOS), an improved electronic voting system for conducting efficient and credible elections in Nigeria that consolidates security as well prevent electoral fraud.

Keywords- Electoral Fraud, Electronic Voting, Rigging, Information Technology, Litigation.

1. INTRODUCTION The paper ballot system, which is still very common in sub- In a democratic society, voting system is no doubt a vital tool that Sahara Africa, Nigeria not an exemption usually, employs allows people to elect the leader of their choice in government or uniform ballots of various stock weights on which the names society. However, voting process in developing countries Nigeria of all candidates and issues are printed. Voter record their inclusive has been faced with lot of challenges associated with choices, in private by marking or thumb print the boxes next traditional voting systems prone to tampering and security to the candidate or issue choice they select and drop the voted infringements. Technology security expert considers ten required ballot in a sealed ballot box. The paper ballot system was first features that characterized a successful voting system. These are adopted in the Australian state of Victoria in 1856 and in the accuracy, convenience, reliability, scalability, flexibility, remaining Australian states over the next several years, where consistency, democratic, timeliness, acceptability, and privacy it became known as the “Australian ballot.” New York which can be achieved today, with the development and became the first American state to adopt the paper ballot for widespread use of information technologies [11][6]. The use of state wide elections in 1889. As of 1996, paper Ballots were computer and information management system has continue to still used by 1.7% of the registered voters in the United States. shaping the society and caused a paradigm shift in decision making They are used primarily as alternative voting system in small process. They are tools for organizing, evaluating and can be used communities and rural areas [8]. to run efficiently, organizational operations [5]. In the mechanical lever voting system, the name of each It is therefore imperative, in this technological age to explore and candidate or ballot issue choice is assigned a particular lever encourage greater use of information technology (IT) in most in a rectangular array of levers on the front of the machine. A forms of service delivery and use as means of transformation of set of printed strips visible to the voters identifies the lever any process, electoral process inclusive. With the rapid assignment for each candidate issue choice. The levers are development of computer technology and cryptographic methods, horizontal in their unvoted position. And when a voter enters efficient electronic voting systems can be employed to replace an the booth and closes the curtain by means of a lever, the inefficient and most importantly error-prone human component in machine unlocks for voting. The titles of all elective offices the present electoral process in the country. are listed on the face of the machine along with the party candidates running for each office. Above each name is a 1.1 Categories of Voting System lever which, when depressed, indicates a vote for that Voting system can be generally categorized into two major types candidate. namely; Traditional Voting System (TVS) and Electronic Voting System (EVS) [7]. The two basic voting systems without electronic means are: (i) paper ballot method and (ii) mechanical lever machine method.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

When the voter pulls the curtain open to leave, the machine The system if espoused will in no doubt bring electoral frauds automatically registers the vote and is cleared for use by the next to the barest minimum as it will eliminate multiple person. Only one candidate for each office is allowed to be registrations and upholds the paradigm of “one voter one selected during this process. If all mechanical connections are vote”[1]. To enhance the efficiency and accuracy of present fully operational during the voting period and the counters are inefficient voting procedures in Nigeria, an improved voting initially set to zero. The position of each counter at the close of the system that consolidate security and fraud prevention can be polls indicates the number of votes cast on the lever that drives it. deployed to help automatically collect and count votes Interlocks in the machine prevent the voter from voting for more electronically. choices than permitted [10]. 2. OVERVIEW OF THE CHARACTER OF In electronic voting system, voter records their secure and secret ELECTIONS IN NIGERIA ballot electronically using modern technological device such as computer and mobile devices where the system can compile and We examine elections and electoral practices in Nigeria in tabulate results automatically. Electronic Voting System (EVS) can four phases. (i). Elections in the period of colonial rule which be described as a voting system by which election data are began with the legislative councils in 1922. (ii). Elections in recorded, stored, and processed electronically, primarily as digital the first years of independence, 1960 – 1965. (iii). Elections information [2]. Electronic voting (also known as e-voting) during the years of military rule and autocracy. The military encompasses both electronic means of casting a vote and counting rulers conducted three elections during their period of rule. the votes. According to Okamoto (1997), e-voting is defined as an These are (a) the elections of 1979, under the first coming of election system that allows a voter to record his or her secure and Obasanjo, (b) the 1992-1993 elections under General secret ballot electronically. The votes data can be digitally stored Babangida and (c) the 1999 elections under General in a storage medium e.g smart card using modern technological Abdusalami Abubakar, and (iv). Elections under civilian device such as computer and mobile devices as a control before regimes from 1983 onwards [4]. Common features of these being sent to a centralized location where the system compile and elections as observed in our survey of elections and electoral tabulate results automatically [9]. practices in Nigeria over these periods have shared a number of common characteristics. They have been particularly Electronic voting technology can include punched cards, optical characterized by massive frauds and intimidations. That is not scan voting systems and specialized voting kiosks including self- to say that there had not been elections conducted in Nigeria contained direct-recording electronic voting machines. It can also with fair results, but most of the time had been flawed. They involve transmission of ballots and votes via mobile devices, were those held in 1959, 1979, 1993 and 1999, while the most computer, or the Internet under dedicated network and security chaotic, violent and disputed were those in 1964 and 1983 system. [13]. The method of voting used in these elections, including that of 2003, 2007 and 2011 was the Open Ballot System in In general, two main types of e-voting can be identified: which the prospective voter goes through a conventional [3][15] process of accreditation, receives a ballot paper from the poll  e-voting which is physically supervised by official and thereafter makes the confidential thumb representatives of governmental or independent electoral impression in favour of the political party or candidate of authorities (e.g. electronic voting machines located at choice in a secret voting compartment before dropping the polling stations); ballot in the box positioned in the open, in the glare of trusted  remote e-voting where voting is performed within the and un-trusted officials, security and party agents. voter's sole influence, though not physically supervised by representatives of governmental authorities (e.g. The modified Open ballot system was adopted in the 1993 voting from one's personal computer, mobile phone, via elections, in which voters filed behind the party symbol or the internet (also called i-voting)) but under a secured photograph of the candidate of choice. Voters were physically network system. counted at the close of polls and the results announced to officials, security and party agents. Although the method is One major benefit of e-voting technology is the speed at which simple and produced what many in Nigeria have often election results could be accurately and automatically counted and described as the fairest and most peaceful elections in the tabulated instantaneously. Other benefits include; reduction in the country, but the winner of the presidential election was never risk of human and mechanical error, movement restriction from officially declared [4].The effect of this since then has one location to another during election since election can take continuously generating issues and crisis in the country. place electronically where voters reside. It also allows easy accessibility for disabled voters. 2.1 Voting System, Electoral Fraud and Crisis in Nigeria Voting system as applicable to Nigeria is still manual. All electoral process starting from registration of voters, accreditation of eligible voters, voting proper and counting of vote cast are done manually.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

However, government have proposed a number of possible The system can be interlinked with a highly secured network methods since voting system commenced in the country to re- that allows for quick and accurate voting electronically. It uses organize and improve the electoral and voting process in the a client/server method which allows voters to cast ballots on country in order to tackle the challenges of traditional approach the client terminal. Each client interfaces with the central characterized by rigging and intimidations [10]. The most recent server, which keeps track of the entire system. The process approach was the introduction of the Biometric Direct Data involves registration, verification, authentication, voting and Capturing Machine (BDDCM), in the last voters’ registration tallying. The intended voter upon registration has a smartcard/ exercise in 2010 in preparation for the 2011 general elections, voters card with his or her bio-data, fingerprint and which would have in a great deal assisted in drastically reducing photograph printed on it to be used during accreditation and fraud and incident of multiple registrations which was usually the election on the Election Day. In the design, the voters’ starting point in election rigging. But this electronic device was database is only accessible to an Independent National edged out of the electoral process during the actual balloting Electoral body official in-charge at a designated centre at because a section of the Electoral Act forbids the use of electronic registration and immediately voters’ data at registration are voting system [12]. Sincerely, to increase the efficiency and validated and locked; only the central authorized administrator accuracy of inefficient voting procedures in Nigeria, an improved has access to the central database/server. voting system that consolidate security and fraud prevention cannot be put off or neglected. Hence the need for an automation In the proposed system, voters can mark their votes directly system, a technique of making a process, or a system operates into an electronic device, using a touch screen, push buttons or automatically, reducing the need for human intervention, errors a similar device when voting. The voting terminals are and fraud. connected via a dedicated network to the central machine, recording votes as they are cast. Alternatively, specially made Electoral fraud and crisis has always been a mile stone to cross for arrangement could be made for people with physical disability many developing countries in sub-Sahara Africa, Nigeria is not an or illiterate, where paper/write-in ballots are permissible, an exemption. By definition, electoral fraud is referred to illegal alphabetic keyboard can be provided to allow voters to interference with the process of an election. It is also called voter cast/write-in votes. These can be received via zip disks or any fraud; the mechanism involved include illegal voter registration, other storage devices. Wherever possible the electoral body intimidation at polls and improper vote counting [14]. Various (INEC) in Nigeria should let experts assist in the forms of statistics could be indicators for election fraud; e.g., exit implementation of the hardware and software, so that another polls which are very different from the final results (i.e empire of bureaucracy is not established to maintain and falsification of election results). Other indicators might be unusual update the technology. With the proposed system, results are high numbers of invalid ballots, overvoting or undervoting, ballot available real-time via an approved medium. Observers and box snatching e.t.c. Generally if the laid down rules for the voters can view the real-time election results from their registration and conduct of elections are not fairly and equitably homes, offices or anywhere in the world using web-enabled obeyed, arbitration will not only bring about the possibility of devices such as PC, laptops, phones or iPAD by simply continued abuse (fraud) and crisis, but the process and the result logging onto the designated website. may be open to prolong legal challenges or litigation. This could undermine the stability of elected body, office, or political stability Meanwhile, the proposed system have been demonstrated of a nation in general [1,12]. over a Local Area Network during the 2011 National Therefore, there is a need for a secure electoral process that will Association of Computer Science Election in Adekunle Ajasin naturally paves way for a free and fair election and has the University, Akungba – Akoko, Ondo State Nigeria, where potential of opening up new opportunities for improved democratic 100% of the elections voting system was carried out using process which of course is important in the context of good computer and results obtained was 97%. (See appendices A- governance and human rights. D) for details. Effort is been made to extend this to other elections in the university. 3. DESIGN OF THE PROPOSED SYSTEM 3.1 Proposed System Components Figure 1 below depicts Network Model Architecture of our 3.1.1 Registration, Verification and Voting Authentication proposed system, Automated Direct Recording Electronic Voting The intended voter is expected to register at designated centres System (ADREVOS) for conducting efficient and credible (wards) by the officials of Independent National Electoral elections in Nigeria. The mechanism is designed to record votes by Commission (INEC), (the body in charge in Nigeria). Multiple means of ballot display with electro-optical components that can biometric traits (e.g fingerprint, face, ear, iris e.t.c) to be be activated by the voters, typically by push button or touch captured to avoid non-universality of traits and a screen; processes votes data by means of a computer program; and smartcard/voters card is issued to respective registered persons records them in memory components. It can produce tabulation of with his or her bio-data, and photograph printed on it. The the voting data and stored them automatically. The system can also identity of a person will be checked using the smart card PIN provide a means for transmitting individual ballots or vote totals to and any of the trait at a confidence level during a central location consolidating and reporting results from accreditation/verification whether he/she is the legitimate precincts at the central location. voter.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The voting machine verifies and authenticates voters by  The machines at the polling boots must be comparing their details with the feature already stored in installed in a secure environment with strong the database under reduced error rate. If the features and access control and monitored by cameras, motion data matches, it accepts/authenticate the voter else the voter sensors, and various other sensor and personnel is denied to vote. This method will enable only legitimate monitoring systems. voters to vote and this can be taken as the first step to avoid  Adequate and strong internal control must be put rigging in the election. The process of voter’s verification is in place because the opportunity for tampering as shown in the form of a flow chart in figure 2 below. applies to persons with inside access and to a lesser extent, outside hackers. Therefore, the 3.2 Voting Process and Counting system must be operated in a secured Upon successful authentication, the system allows environment. electorate to vote. There appears a screen on which all the  The machines should be delivered to the polling parties’ candidate’s names with their corresponding locations in a manner that prevents anyone from positions and symbols appeared. The user is to select tampering with them without it being choice by selecting a radio button in front of each party or immediately evident to election poll workers. using touch screen. After an option is selected, the voter is prompted for a confirmation. And once a vote has been cast  The inter-communication network must be for a particular candidate and position, the voting page dedicated with high signal strength. scroll off and disallows the voters for any illegal attempt to  Standby security agents to maintain orderliness at vote twice. Based on the option selected, the vote count of the voting boots. Only registered voters should be the particular candidates gets incremented automatically. allowed to vote, logically each voter cannot vote See appendix A and B. Once a vote is cast the voter is more than once. given a receipt or counterfoil for his/her vote, time of vote and where the vote was cast. The proposed system is designed desirably of having an electronic data backup both on-site and off-site. Having 3.3 Basic Security for the proposed system such as parallel simultaneously recording hard drives, other The following specific security measures must be critically backup equipment and an uninterruptible power supply taken into consideration during implementation of the (UPS). This is always desirable in any system using proposed system: computer equipment, and it is especially important in the case of electronic voting machines because of the extreme public importance of not losing votes and having uninterrupted service on voting days.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Network Model Architecture of the Proposed Mechanism

Secured Multimodal Biometric Registration, Verification Internet/Network & Voting Centres Voters

Location

1,2….n

….

Voters’Voters’ Database Database Voters Locationlocation INEC 1,2..n Voters 1,2…n Pooling boot z Pooling boot x Official Location 1,2…n

Generates Voters’ INEC Sever PIN & Smart Cards INEC Serveir

Automated online voting capturing

Automatic votes counting and tabulation

Online Information e.t.c. INEC

System Administrator

Figure 1: ADREVOS Network Model Architecture

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Voter’s Verification Components and Flowchart

Start

Enter Smart card

pin/ use captured biometric traits

NO Message Prospectiv “Record not Biometri e voters c found” verified? Database Voters denied to vote

YES

Voters allowed to vote

INEC Server/

Vote Database Stop

Figure 2: Voter’s Verification Component and Flow of control

4. SUMMARY 5. CONCLUDING REMARKS

The best way to protect the electorate from electoral fraud is Electricity supply, is one major challenge against the use of to have an election process which is completely transparent to any electronic machine, it is necessary to posit that these all voters, devoid of fraud from nomination of candidates devices can be implemented using a 6 volts battery to sustain through casting of the votes and tabulation. A key feature in a unit of the equipment for up to forty-eight hours. While not ensuring the integrity of any part of the electoral process is a under any illusion, the adaptation of ADREVOS will strict chain of custody. Automated Direct Recording Voting drastically reduce if not completely erase all cases of electoral System (ADREVOS) proposed in this publication has a lot of fraud in the country. In addition, the present electoral reform benefits over present voting system in Nigeria. It’s accuracy, in the country should give room to accommodate this new convenience, reliability, and efficiency has a lot to offer in the move. I am therefore, convinced that this new process will electoral process of the country and other developing remove major lapses observed at previous elections and make countries. There is no doubt that, if the proposed electronic it more difficult to perpetrate fraud and subvert the legitimate voting system is rightly and fully implemented have the desire of the voting public. potential to improve the present conventional voting procedures by providing added convenience and flexibility to the voter and electoral bodies. Moreover, since paper work may not be required as such in the new system; it will be more economical and counting of votes will be instantaneous, direct, accurate and automatic. In addition to this, the chances of people voting more than once, dead people voting e.t.c will be drastically reduced because of the expert system built in.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

6. REFERENCES [8]. Manish K, Suresh K., Hanumanthappa M., Geetha D.E (2009). Secure Mobile Based Voting System. [1]. Andrew R., and Paul B. (2003): Bringing Confidence to Dept. of Computer Science and Applications, Central Electronic Voting, EJEG, Vol 1, Issue 1. August 2008. College Campus, Bangalore University, Bangalore- Available online http://www.ejeg.com 560 001, India

[2]. Bellis, M. (2011) The History of Voting Machines. [9]. Okamoto T. (1997)”Receipt free- electronic voting www.about.com. Retrieved 11th July, 2012. schemes or Large scale election”, Proc., of workshop on security protocols. [3]. Buchsbaum, T. (2004). "E-voting: International developments and lessons learnt". Proceedings of [10]. Paul et al (2008): Voting Technology: The Not-So- Electronic Voting in Europe Technology, Law, Politics Simple Act of Casting a Ballot, Published by the and Society. Lecture Notes in Informatics. Workshop of Brookings Institution .Retrieved August 15 the ESF TED Programme together with GI and OCG. 2009..http://www.brookings.edu/press/Books/2007/vot ingtechnology.aspx [4]. Iyayi. F. (2004): The Conduct of Elections And Electoral Practices In Nigeria. Being Paper delivered at the NBA [11]. Rubin A. D., (2002), “ Security Considerations for Conference in Abuja on 24th August, 2004. Department Remote Electronic Voting”, Communications of the of Business Administration, Faculty of Social Sciences, ACM, 45(12):39–44, December, 2002 University of Benin, Benin City, Edo State, Nigeria. [12]. Saltman, R. (2012): Effective Use Of Computing [5]. Iwasokun, G B, Alese, B. K. Thompson, F.B, and Technology In Vote-Tallying. Nist. Remote Voting Aranuwa F.O. (2012) Computer Model Analysis of the Technology, chris backert e-government consulting. Performance Of ICT in the Nigerian Universities. International Journal of Advanced Research in Computer [13]. The Nigerian Elites Forum, (2011): History of Science and Software Engineering Research Paper. democratic elections in Nigeria. Volume 2, Issue 1, January 2012 ISSN: 2277 128X. http://www.nigerianelitesforum.com/ng/elections-and- Available online at: www.ijarcsse.com. electoral-issues/8441-history-of-democratic-elections- in-nigeria.html#ixzz22je8j0T4 [6]. Jefferson, D, Rubin, A D, Simons, B. and Wagner. D [14]. Wikipedia, (2012): The Free Encyclopedia. Retrieved (2004) A Security Analysis of the Secure Electronic th Registration and Voting Experiment (SERVE), ETS 300 on 25 July, 2012. 506. Security aspects (GSM 02.09 version 4.5.1), Digital cellular telecommunications system (phase 2), 2000. [15]. Zissis, D and Lekkas S (2011). "Securing e- Government and e-Voting with an open cloud [7]. Jones D. W., (2001), “A Brief Illustrated History computing architecture". Government Information of Voting” Department of Computer Science, Quarterly 28 (2): 239–251. University of Iowa, USA. doi:10.1016/j.giq.2010.05.010.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

APPENDICES

Appendix A: Prototype of Registration and Verification Process Interface

Appendix C: Prototype of Ballot Screen Display of Candidates

Appendix B: Prototype of Bio-Data Registration Process Interface

Appendix D: Prototype of Tabulated Election Results

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Modeling Variation of Waiting Time of Distributed Memory Heterogeneous Parallel Computer System Using Recursive Models

O.E. Oguike, M.N. Agu & Echezona, S.C. Department of Computer Science, University of Nigeria, Nsukka, Nigeria. [email protected], [email protected], [email protected]

ABSTRACT In a heterogeneous parallel computer system, the computational power of each of the processors differs from one another. Furthermore, with distributed memory, the capacity of the memory, which is distributed to each of the processors, differs from one another. Using queuing system to describe a distributed memory heterogeneous parallel computer system, each of the heterogeneous processors will have its own heterogeneous queue. The variation of waiting time of heterogeneous parallel computer system with distributed memory needs to be modeled because it will help designers of parallel computer system to determine the extent of variation of the waiting time. It will also help users to know when to realize minimum variation of the waiting time. This paper models the variation of the waiting time of distributed memory heterogeneous parallel computer system using recursive models. It also uses the statistical method of Z-Transform to verify and validate the recursive model.

Keywords- Heterogeneous parallel computer; distributed memory; parallel computer system; queuing network; variation; recursive models; waiting time; Z-Transform.

I. INTRODUCTION

A heterogeneous parallel computer system is one in which the computational power of each of the processors differs from one another. With distributed memory, it means that each of the cpu queue heterogeneous processors has its own memory. Describing the Parallel system using queuing network, each of the processors has its own Pocessors queue. With a round robin scheduling algorithm, processes can be scheduled to the various parallel processors, whenever a process cpu queue needs to perform an I/O operation, it joins the appropriate I/O queue. Therefore, the queuing network of a heterogeneous parallel computer system consists of parallel processors, parallel processor cpu queue queues, I/O processors and I/O queues. Suppose there are n different parallel processor queuing systems and k different I/O queuing systems. A queuing system in this context is defined as a cpu queue processor, together with its own queue.

We assume that the various queues are finite [1, 2, 3, 4] i.e. there is a limit to the number of jobs that can be admitted into the queues, and negligible communication overhead. Suppose X , X , X , … , I/O queue 1 2 3 Xn , Xn+1, Xn+2, Xn+3, …, Xn+K are the maximum number of processes that can be admitted into the respective queues. We assume that processes arrive at the various queues according to I/O queue Poisson distribution, and they are serviced according to Exponential distribution [5, 6]. Figure 1 illustrates a model of the queuing network of a heterogeneous parallel computer system with distributed memory. I/O queue

I/O processors

Figure 1: Queuing network of a heterogeneous parallel computer system with distributed memory.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

There are different performance metrics of a parallel computer Therefore, recursive models can be used to efficiently determine system that can be modeled, however, for distributed memory the exact convergence of any series used in modeling the variation heterogeneous parallel computer system, variation of waiting time waiting time of a distributed memory parallel computer system. is an important performance metric that needs to be modeled. This is because the various computational resources and processes are 3. DEVELOPING THE RECURSIVE MODELS heterogeneous, therefore there is need to measure the extent of variation between the heterogeneous computational resources and The recursive model was developed for one queuing system; processes. afterwards, it was generalized to consider all the queuing systems

of the queuing network. As a result, the following models have 2. LITERATURE REVIEW AND LIMITATION OF been developed for one queuing system and for all the queuing CURRENT TECHNIQUE systems of the queuing network.

Queuing approach has been used extensively in the literature to A. Models Based on a Queuing System model the performance of computer systems. However, this has The following models have been developed for one queuing been done in different ways and for different models of computer system systems. In [20], the authors used a recursive computation Recursive Probability Density Function of the Number of approach to solve the steady state equations, thereby leading to the Processes in Queuing System. modeling of the various performance metrics of a multi-terminal system that is subject to breakdown. Furthermore, the author in Let X denotes the maximum number of processes that can be [24] used a rigorous approach to model the performance of i in the ith finite queuing system at any time [12, 13, 14]. Suppose heterogeneous parallel computer system without introducing any constraint on the kind of interconnection between the the arrival rate, xi when xi processes are in the ith queuing system heterogeneous nodes. Furthermore, in [24], systems with the same of the queuing network be described as: interconnection speed were considered when modeling the  , x  0,1,2,,3,...X 1 performance of heterogeneous parallel computer system. The i i i x i   (1) authors in [25] looked at alternative ways of measuring the 0, otherwise performance of heterogeneous parallel computer system, by Since the various processors are heterogeneous, therefore, it modeling linear speed and linear efficiency using simulation- implies that the departure rate will vary, which can be described as: modeling techniques.

 , x 1,2,3,4,..., X In [26], the author showed that Little’s formulae could be  i i i (2)  x   universally applicable, if properly interpreted to take account of i 0, otherwise state-varying entrance rates, batch arrivals, and multiple customer  classes. In [27], the author confirmed that Little's formula could be applied to very general queuing systems (not just M/M/1), even Using the steady state probability as stated in [7, 16] the whole networks! The authors in [28] considered a new performance probability that xi processes will be in the ith queuing system is metric, variation of the computing power as a unique performance metric that is ideal for a heterogeneous network of workstations,  x though an approach different from queuing approach was used to i P0i , x  X i P  (3) do this. In [29], analytic models were used to model the xi  performance of computer intensive applications of parallel 0, otherwise computers, while [30] used recursive models only to evaluate the performance of compute intensive application of a parallel The utilization factor for the ith queuing system,  is defined as: computer system. In [31], recursive models were used to evaluate i various performance metrics of heterogeneous parallel computer i system with distributed memory; however, variation was not part of . To obtain the value of P0i in equation (3), we sum all the the performance metric modeled. In [33], the authors used  i recursive model to model the variation of the average number of probabilities for the ith queuing system and equate it to 1. processes in the system, though the developed recursive models were not validated. This implies that: X Though analytic queuing method has been used in literature [29,32] i . (4) to model the performance metrics of various computer queuing  Pxi  1 models, however, one limitation of the analytic method is its xi 0 inability to efficiently determine the exact convergence of some mathematical series that are used in modeling variation of waiting From equations (3) and (4), it implies that: 2 3 4 X time of distributed memory, heterogeneous parallel computer P0i+iP0i+i P0i + i P0i + i P0i + … + i iP0i = 1. (5) system. Therefore, there is the need for another model, rather than analytic model. The use of efficient linear recursive model [9] can Factorizing equation (5), it implies that 2 3 X efficiently model the variation of waiting time of a distributed P0i (1 + i + i + i + … + i i) = 1. (6) memory, heterogeneous parallel computer system.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Recursive model can be used to show where the series in (6) 1 converges. The recursive model is given below as: Poi = (10) Sum1(Xi , i ) 1 , X = 0 Using equation (10) in equation (3), we have the following:

Sum1(X,  ) = (7)   xi i , x  X  i (11) Pi x  Sum1(X i ,  i ) i  Term1(X, ) + Sum1(X-1, ), X  0 0,Otherwise

The recursive algorithm that can be used to implement the recursive model in equation (7) is given below as: Equation (11) is the probability density function that models the probability that xi processes will be admitted in the ith queuing Double sum1(int X, double  ) system. 1. Request data 1.1 Request X Average Number of Processes in One Queuing System. 1.2 Request  Furthermore, the average number of processes in the ith 2. Determine Sum1 queuing system (i.e the queue and the processor) can be described 2.1 Sum1 = 1 if X = 0 else Sum1=Term1(X,  )+Sum1(X-1,  ) statistically as expectation of xi , where xi is the random variable 3. Display Sum1 that denotes the number of processes in the ith queuing system. This can be written as Term1(X, ) is the recursive model that determines the xth term of the series in (6), it is given as: X i E( x ) = x P . (12) i  i i xi 1, Xi = 0 xi 0

Using equation (11) in equation (12), we obtain the following: x Term1i(Xi,  i ) = (8) X i i  xi i  E(x )    (13) i    x 1 Sum1(Xi , i ) i * Term1i(Xi-1,  i ), Xi 0 i  

The recursive algorithm that can be used to implement the Equation (13) can be simplified as: recursive model in equation (8) is given below as:  1  2 3 X 1 Double Term1(int X, double ) E(x )    1 2  3  4  ...  X  i (14) i  Sum1(X ,  )  i i i i i i 1. Request data  i i  1.1 Request X 1.2 Request A recursive model has been used in [30,31] to determine the convergence of the series in equation (14). The recursive model is

2. Determine Term1 called Sum2i(Xi,  i ), and it is given as: 2.1 Term1 = 1 if X = 0 else

Term1=  *Term1i(Xi-1, ) 1, Xi = 1

3. Display Term1 (15)

Using equation (7) in equation (6), we obtain the following:

Poi Sum1(Xi,  i ) = 1 (9) Term2i(Xi)*term1i(Xi-1,  ) + Sum2i(Xi-1,  i ), Xi 1 i

The recursive algorithm that can be used to implement the

recursive model in equation (15) is given below as: Solving for Poi in equation (9), we obtain the following:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

 Double Sum2(int X, double  ) i i  , i 1,2,3,..., n  k (18) 1. Request data 0i 1.1 Request X 1.2 Request  Variation of Average Waiting Time in all the Queuing 2. Determine Sum2 Systems of the Queuing Network 2.1 Sum2 = 1 if X = 1 else Suppose xi is the random variable that denotes the number of Sum2= Term2 (X )*term1 (X -1, ) + i i i i  i processes in the ith queuing system. Therefore, the average or mean processes in all the queuing systems of the queuing network can be Sum2i(Xi-1,  ) i defined as 3. Display Sum2 nk x Term2i(Xi) is given as:  i Y  i1 1, Xi = 1 n  k (19)

Term2 (X ) = Little’s formula can be used to related equation (19) to the average i i waiting time in all the queuing systems of the queuing network. (16) Therefore, using the constant of proportionality of Little’s formulae [7], we can establish a relationship between the average number of processes in all the queuing systems and the average waiting time 1 + term2i(Xi-1), Xi  1 in all the queuing systems of the queuing network., as shown in equation (20). The recursive algorithm that can be used to implement the recursive model in equation (16) is given below as:  nk  Double Term2(int X)  x   i (20) 1. Request data i1 1 Ws    1.1 Request X   n  k e 2. Determine Term2   ff 2.1 Term2 = 1 if X = 1 else  

Term2= 1 + term2 (X -1) i i The constant of proportionality, 3. Display Term2

nk term1i(Xi, ) is the recursive model in equation (8). e  ffi Therefore, using equation (15) in equation (14), we obtain:   i1 (21) e ff n  k  Sum2i (X i , i )  E(x )    (17) i  Sum1 (X ,  )  Using equation (20), we can take the variance of the average  i i i  waiting time as:

 nk   B. Models Based on The Whole Queuing Network.  x   Having developed the models for the performance metrics of one  i 1 VAR(Ws)  VAR i1   (22) queuing system, these models can be extended to the whole    queuing systems of the queuing network of a heterogeneous n  k e ff    parallel computer system. It is necessary to define ii as the    probability that a process will join the ith queue after each cpu Using one of probability theory laws in [23], we obtain: burst, and 0 as the probability that the execution of a process has been completed. Arrival of processes into the various parallel 1 nk processor queues can come from the outside world or from the VAR(Ws)  VAR x 2 2   i  various I/O queues or from the particular parallel processor, at the (n  k) e i1 ff (23) expiration of the time quantum for that process. Let i be the rate of arrival of processes into the ith queuing system, and  , the rate From [23], the variance can be defined statistically as: 2 2 of arrival of processes from the outside world.. VAR(x )  E(x )  (E(x )) (24) Under the steady state, when we consider the queuing network, the i i i overall utilization factor has been defined in [31] as: Simplifying equation (24) further, we obtain:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

2 2 Double Term3(int X) E(xi )  xi Pi (25)  xi 1. Request data 1.1 Request X Using equation (11) in equation (25), we obtain: 2. Determine Term3 2.1 Term3 = 1 if X = 1 else 2 x X i i Term3= (2*Xi-1) + Term3i(Xi-1) 2  x   E(x )   i i  (26) 3. Display Term3 i    x 1 Sum1 (X ,  ) i  i i i  The recursive model that determines the xth terms of sequence2 has been developed in equation (8). Therefore, combining Simplifying equation (25), we obtain: equation (30) and equation (8), the series in equation (29)

converges to this recursive model, called Sum3 (X ,  ), which is   i i i 2 1 2 2 2 2 3 2 4 2 X i (27) E(xi )   1 i  2 i  3 i  4 i  ...  X i i  shown below as:  Sum1 (X ,  )   i i i  1, Xi = 1 Simplifying equation (27) further, we obtain:

  2 1 2 3 4 2 X i E(xi )   1 i  4i  9i  16i  ...  X i i  (31)  Sum1 (X ,  )   i i i  (28)

Term1i(Xi-1,  i )*Term3i(Xi) + Sum3i(Xi-1,  i ) ,Xi 1 Factorising equation (28), we obtain: The recursive algorithm that can be used to implement the  1  recursive model in equation (31) is given below as: 2   2 3 2 X i 1 E(xi )   i 1 4i  9i 16i  ...  X i i  Sum1 (X ,  ) Double Sum3(int X; Double  i )  i i i  (29) 1. Request data 1.1 Request X The convergence of the series may not be efficiently determined 1.2 Request  analytically; therefore we seek for its convergence using recursive 2. Determine Sum3 models. The same approach used earlier can be used to determine 2.1 Sum3 = 1 if X = 1 else the convergence of the series,

2 3 2 Xi 1 Sum3= Term1i(Xi-1, )*Term3i(Xi) + 1 4i  9i 16i  ...  X i i . The series can be considered as two sequences, which are: sequence1 = 1, 4, Sum3i(Xi-1, ) 9, 16, …, X2, while the other sequence is: sequence2 = 3. Display Sum3

2 3 X i 1 1, i , i , i ,..., i . The recursive model Therefore, using equation (31) in equation (29), we obtain: that can be used to determine the xth terms of sequence1 can be obtained by adding 2X-1, which is the common difference   Sum3(X ,  )  between the xth term and the (x-1)th term, to the (x-1)th term of 2 i i i  E(xi )    (32) the sequence. The recursive model can be represented as shown    Sum1(X i , i )  below in equation (30), as: Using equations (17) and (32) in equation (24), we obtain: 1, Xi = 1

2   Sum3(X ,  )    Sum2(X ,  )  Term3i(Xi) = (30) VAR(x )   i i i    i i i  (33) i  Sum1 (X ,  )   Sum1 (X ,  )   i i i   i i i 

(2*Xi-1) + Term3i(Xi-1), Xi  1

Therefore, using equation (33) in equation (23), we obtain: The recursive algorithm that can be used to implement the recursive model in equation (29) is given below as:

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

2 The statistical method of Z-transform can be used to validate the  n  k  1   Sum3(X ,  )    Sum2(X ,  )   recursive models. The Z-transform for the ith queuing system, VAR(Ws)    i i i    i i i   (34) 2 2        using the statistical generating function is given as: (n  k)  i 1  Sum1 (X ,  ) Sum1 (X ,  )  e ff  i i i   i i i     x GXi(z) = E(z ) (35) Equation (34) models the variation of the average waiting time in all the queuing systems of the queuing network. Therefore, using equation (35), the variance of the ith queuing system can be expressed in terms of the z-transform as in [25]:

4. VERIFYING AND VALIDATING THE MODELS USING 2 Z-TRANSFORM  G Xi (z)  GXi (z)  VAR( X i) = z = 1 +   z = 1 - (z) 2  (z)  Model verification and model validation are essential parts of   2 model development that will help to assess the quality of the   developed models. If a model is not verified and validated it cannot G Xi (z)   z = 1 (36) be assured of quality, therefore, it can be sent back to the drawing  (z)  board. Model verification is done in order to ensure that the   simulation algorithm i.e. algorithms used to implement the models on the computer are correct and the simulation programs i.e. model Simplifying equation (36) further, using the analytic model for the implementation programs are correctly programmed. Model probability density function for i  1 , as stated in [33], we verification eliminates every error that may occur when obtain the following: implementing the models on the computer. On the other hand, model validation aims at making the model address the right x problem, address accurate information about the system being X i   i (1  )  modeled. Model validation compares the results of the simulated xi  i i  GXi (z)  z (37)   X i 1  models with the results of a real system. Therefore, model x 0 1  validation tries to establish if the model is an accurate i  i  representation of the real system. However, due to one reason or the other, it may not be easy sometime to obtain results of the real Simplifying further, we obtain the following: system, in such a situation, expert knowledge can be used to X  (1  )  i determine if the qualitative data from the simulated model is valid i xi xi GXi (z)    z i or invalid [35].  Xi 1  1  x 0  i  i (38) The authors in [34] argued that though quantitative comparison will provide the basis for validation, however, it can miss the qualitative Simplifying further, we obtain the following: discrepancies or agreements that human are capable of detecting. One of the ways they suggested that can be used to detect such X discrepancies or agreements is through visualization. Visualization,  (1  )  i G (z)   i  (z ) xi according to them helps to map numerical data into graphical Xi X 1  i structure that human can more readily understand. This graphical  i  1 i xi 0 display of the results of the simulated model or the system behavior (39) will help us to determine if the model is valid or invalid. Simplifying further, we obtain the following: Furthermore, [34] pointed out that quantitative comparison is needed to make finer distinctions between behaviors that agree in their basic form, but qualitative comparison can help to eliminate  (1  )  1 (z ) X i 1  models that are not in the right ballpark [34]. Sometimes a  i   i  GXi (z)    validated model can be used to validate another model by 1  X i 1  (1 (z ))  comparing qualitative and quantitative data of the two models.  i  i  (40)

Therefore, taking the first derivative of equation (40), with respect to z, and initializing z to 1, we obtain the following:

X 1 X G Xi (z)  (1  i ) 1 X   (X 1)      (41) z  X i 1  2  1  i  (1 ) 

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Simplifying further, equation (41) reduces to: 6. RESULTS OF THE SIMULATION

X 1 X The results of the simulation have been analyzed to determine how G Xi (z)   1 X   (X 1)      (42) variation of the waiting time changes as a particular parameter z 1  X i 1  (1 )  varies, while other parameters remain constant [10]. Table 1 and  i   figure 2 show the result of the simulation, suppose the probability of a process leaving the system is known to be 0.2 and the Furthermore, taking the second derivative of equation (40), probabilities that a process will join the first and second queues are with respect to z and initializing z to 1, we obtain the following: 0.775 and 0.025, respectively. Suppose the first processor is a high- speed processor with high departure rate of 30, while the second 2 processor is a low speed processor with a low departure rate of 10.  GXi (z)  (1 i )  u  v  Suppose the maximum number of processes to be allowed into first     (43) queue is 20, while maximum number of processes to be allowed 2  X i 1  4  (z) 1 i  (1 )  into the second queue is 5. The experimental trials were carried out several times, in each trial, the arrival rate was changed, and the Simplifying further, u and v are given as: corresponding variation was obtained as the result of the simulation.

2 X i 2 2 X i 1 X i 1 u  1 i  Xi (Xi 1)i  Xi i  Xi i  (44) TABLE 1: RESULT OF VARIATION AGAINST ARRIVAL RATE

AR V.From Model V.From Z-Transform Xi 2 Xi 1 X i 1 v  2i 1 i Xi  X i i  i  i  (45) 3 0.00186 0.00186

4 0.00221 0.00221 Therefore, using equations (44) and (45) in equation (43), and using equations (43) and (42) in equation (36), we obtain the z- 5 0.00323 0.00323 transform model for the variation of waiting time in the ith queuing system. Furthermore, the z-transform can be used to obtain 6 0.00578 0.00578 variation of the average waiting time in all the queuing systems of 7 0.00978 0.00978 the queuing network, as shown in equation (46) below. 8 0.00999 0.00999 1  nk  9 0.00623 0.00623 VAR(Ws)   VAR(X ) (46) 2  i 2 10 0.00341 0.00341 (n  k) e ff  i1  11 0.002015 0.002015 However, the z-transform cannot be used to effectively validate 12 0.001332 0.001332 the recursive model for the isolated case when i  1. Key to the Table: AR: Arrival Rate. 5. METHODOLOGY V.from Model: Waiting Time Variation, using Recursive models. This paper has used recursive models to model the variation of V.from Z-Transform: Waiting Time Variation, using Z-Transform. waiting time of distributed memory, heterogeneous parallel computer system. A queuing approach, with finite queues has been Waiting Time used to achieve the above aim, with parallel processors depicting Variation Against parallel servers. The statistical method of probability density Arrival Rate function and other probability theory concepts have used [15, 23]. 0.012 A novel method of deriving the recursive model that determines the 0.01 xth terms and the convergence of important mathematical series 0.008 have been used to develop the recursive models. The simulation of 0.006 the models on the computer has been done using Java 0.004 programming language and the statistical regression/trend line analysis has been used to analyze the results of the simulation [11]. 0.002 0 The simulated recursive models have been validated using 0 5 10 15 statistical method of Z-Transform. Arrival Rate

Variation Time Waiting

Figure 2: Variation Against Arrival Rate

The undulating nature of the result shows the various points where minimum variations and maximum variation can be realized.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Furthermore, table 2 and figure 3 show the simulation results as TABLE 3: RESULT OF VARIATION AGAINST DEGREE OF we keep the following input parameters constant, the probability MULTIPROGRAMMING that a process will leave the network is 0.2, the probabilities that a process will join queue 1 and 2 are 0.775 and 0.025, respectively, TMP V. from Model V. from Z-Transform while the departure rates for processor 1 and 2 are 30 and 10, 8 2.82E-04 2.82E-04 respectively, and the maximum number of processes in queue 1 and 2 (degree of multiprogramming for the two queues) are 20 and 13 3.09E-04 3.09E-04 5, respectively, and the arrival rate from the outside world is 4 (for 18 3.13E-04 3.13E-04 non-compute intensive applications) and 30 (for compute intensive applications). By changing the degree of multiprogramming 23 3.14E-04 3.14E-04 (maximum number of processes in the system) for the two queues of a two-processor parallel computer system, we obtain the 28 3.14E-04 3.14E-04 corresponding variations shown in table 2 and figure 3 for non- 33 3.14E-04 3.14E-04 compute intensive applications. 38 3.14E-04 3.14E-04 TABLE 2: RESULT OF VARIATION AGAINST DEGREE OF 43 3.14E-04 3.14E-04 MULTIPROGRAMMING 48 3.14E-04 3.14E-04 TMP V. from Model V. from Z-Transform 8 0.0010662 0.0010662 Table 3 and figure 4 show the results of the waiting time variation against the total maximum number of processes for compute 13 0.0015138 0.0015138 intensive applications, i.e. when the overall utilization factor is 18 0.0020149 0.0020149 greater than 1. The behavior of the waiting time variation is the same for both compute and non-compute intensive applications. 23 0.0021338 0.0021338 Waiting Time Variation Against Total Maximum 28 0.0021958 0.0021958 Number of Processes 33 0.0022076 0.0022076 38 0.0022127 0.0022127 3.20E-04 3.10E-04 43 0.0022135 0.0022135 3.00E-04 48 0.0022138 0.0022138 2.90E-04 Key to the Table: Variation TMP: Total Maximum Number of Processes. Time Waiting 2.80E-04 V.from Model: Waiting Time Variation from Model. 0 10 20 30 40 50 60 V.from Z-Transform: Waiting TimeVariation from Z- Total Maximum Number of Processes Transform.

Waiting Time Variation Against Total Figure 4: Variation Against the Degree of Multiprogramming Maximum Number of Processes In a similar manner, as we keep the following input parameters constant, probability of a process leaving the network is 0.2, while 0.0025 the probability of a process going to queue 1 and 2 is 0.4, the 0.002 arrival rate from the outside world is 5. The maximum number of 0.0015 0.001 processes that can be in queue 1 and 2 are 15 and 14, respectively. 0.0005 By changing the departure rates of the two processors, we obtain 0 the corresponding variations of the waiting time, as shown in table Waiting Time Variation Time Waiting 0 10 20 30 40 50 60 4 and figure 5. The result shows that the behavior of the waiting Total Maximum Number of Processes time variation for compute intensive applications, i.e. when the overall utilization factor is greater than 1 is different from the behavior of the waiting time variation for non-compute intensive Figure 3: Variation Against the Degree of Multiprogramming applications, i.e. when the overall utilization factor is less than 1. From the results in table 3 and figure 4, increasing the speed of the From the results in table 2 and figure 3, it can be seen that for non- processors for compute intensive applications will lead to a compute intensive applications, where overall utilization factor is corresponding increase in the waiting time variation. On the other less than 1, as the total maximum number of processes in all the hand, increasing the speed of the processors for non-compute queues increases, the waiting time variation increases, but intensive applications will lead to a corresponding decrease in the afterwards, it remains constant. waiting time variation.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[5] Cathy H. Xia, Zhen Liu., Queueing systems with long-range Waiting Time Variation Against Total Total dependent input process and subexponential service time. Departure Rates Proc. ACM SIGMETRICS international conference on Measurement and modeling of computer systems,(2003). [6] Shanti Subramanyam, Performance Modelling of a J2EE 0.03 Application to meet Service Level s, Agreement, Proc. 0.025 International Conference of Computer Measurement Group, 0.02 (2005) 0.015 0.01 [7] Hamdy A. T.,. Operation Research: An Introduction,

0.005 Prentice-Hall of India, (1999). Variation 0 [9] Ivan Stojmenovic; Recursive Algorithms in Computer WaitingTime 0 10 20 30 40 50 Science Courses : Fibonacci Numbers and Binomial Coefficients; IEEE Transactions on Education; Vol. 48, No. 3 Total Departure Rates [10] Arjan J.C. van Gemund; Performance Modelling of Parallel Systems: An Introduction. Figure 5: Variation Against Total Departure Rate [11] Justyna Berlinska, The Statistical models of parallel applications, Annales UMCS Informatica, (2005). TABLE 4: RESULT OF VARIATION AGAINST DEPARTURE RATE [12] Arranchenkov, K.E., Vilchersky, N.O., Shevlyakor, G.L TDR1 TDR2 WTVM WTVZT Priority queueing with finite buffer size and randomized push-out; 1 3 0.011495 0.011495 mechanism. Proc. of ACM SIGMETRICS international 3 5 0.010178 0.010178 conference on measurement and modeling of computer 5 7 0.015133 0.015133 systems.; (2003). 7 9 0.023701 0.023701 [13] Abunday, B.D., and Khorram, E. The finite source queueing 9 11 0.025995 0.025995 model for multiprogrammed computer systems with different 11 13 0.018658 0.018658 CPU times and different I/O times. Acta Cybern. 8, 4 , (1998) 13 15 0.010034 0.010034 [14] J. Sztrik; Finite-Source Queueing Systems and their Applications: A Biliography; 15 17 0.005631 0.005631 [15] Trivedi K. Shridharbhai, Probability and Statistics with 17 19 0.003593 0.003593 Reliability, Queuing and Computer Science Applications, 19 21 0.002543 0.002543 John Wiley & Sons Inc., (2002). 21 23 0.001934 0.001934 [16] Per Brinch Hansen. Operating System Principles. Prentice- Key to the table: Hall of India Private Limited, (1990). a DRP1 Departure Rate for Processor 1 [20] J. Sztrik and T. Gál A recursive solution of a queueing DRP2: Departure Rate for Processor 2 model for a multi-terminal system subject to breakdowns; WTVM: Waiting Time Variation from Model Performance Evaluation Volume 11, Issue 1, Published by WTVZT: Waiting Time Variation from Z-Transform Elsevier, (1990). [23] Robert V. Hogg and Allen T. Craig; Introduction to Mathematical Statistics; Macmillan Publishing Co. Inc.; 7. SUMMARY AND CONCLUSION (1978). [24] Andrea Clemantis, Angelo Corana; Modelling Performance This paper has been able to model the variation of a waiting time of of Heterogeneous Parallel Computer System; Journal of Parallel Computing, Volume 12, Issue 9, Elsevier; pages heterogeneous parallel computer, using recursive models and 1131-1145; (1999). queuing approach. The models have been simulated on the [25] E. Post, H.E. Goosen; Evaluating the Parallel Performance of computer using Java programming language and validated using a Heterogeneous System statistical Z-Transform method, the results of the simulation have [26] Beutler, F; Mean sojourn times in markov queuing network: been analyzed in order to determine when to realize minimum Little’s formula revisited; IEEE Transaction on Information variation. Theory; Volume 29, Issue 2, page 233-241; (2003). [27] Ken Vastola; REFERENCES http://networks.ecse.rpi.edu/~vastola/pslinks/perf/node46.htm l [28] Xiaodong Zhang, Yong Yan; Modeling and Characterizing [1] Henry H. Liu and Pat V. Crain, An Analytic Model for Parallel Computing Performance on Heterogeneous Network Predicting the Performance of SOA-Based Enterprise of workstations; Proceedings of the 7th IEEE Symposium on Software Applications, Proc. International Conference of Parallel and Distributeed Processing (SPDP ’95) 1063- Computer Measurement Group, (2004). 6374/95 $10.00 © 1995 IEEE [2] S. Balsamo et al, A Review of Queueing Network Models [29] O.E. Oguike et al; Modelling the Performance of Computer with Finite Capacity Queues for Software Architecture Intensive Applications of Parallel Computer System; Proc. Of Performance Prediction, (2002). IEEE 2nd International Conference on Computational [3] Catalina M. Liado et al, A Performance Model Web Service, Intelligence, Modeling and Simulation; (2010). Proc. International Conference of Computer Measurement [30] O.E. Oguike et al; Evaluating the Performance of Parallel Group, (2005). Computer System Using Recursive Models; Proc. Of IEEE [4] Rosselio, J et al, A Web Service for Solving Queueing 4th UKSim European Modeling Symposium; (2010). Network Models Using PMIF. www.perfeng.com/paperndx.htm, (2005).

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[31] O.E. Oguike et al; Evaluating the Performance of Heterogeneous Distributed Memory Parallel Computer nd System Using Recursive Models; 2 IEEE International Conference on Intelligent Systems, Modeling and Simulation; (2011). [32] Leonard Kleinrock, Queueing Systems Volume 1 and 2, John Wiley & Sons, (1975). [33] O.E. Oguike et al; Modelling Variation of a Performance Metric of Distributed Memory Heterogeneous Parallel Computer System, Using Recursive Models; In proc. of 3rd IEEE International Conference on Computational Intelligence Modeling and Simulation; (2011). [34] Bernard P. Zeigler et al; Theory of Modelling and Simulation; Elsevier; (2000) [35] Cor van Dijkum et al; Validation of Simulated Models; Siswo Publication 403, Amsterdam, (1999)

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Usable Authentication Schemes: A Critique

1O. O. Ayannuga Ph.D & 2O. N. Lawal Computer Technology Department Yaba College of Technology Yaba-Lagos, Nigeria. [email protected], [email protected]

ABSTRACT Usable authentication was given birth to as a result of memorability shortcoming; that is, low memorability of passwords by the users of authentication systems. To improve on the flaws of text based authentication schemes, usable authentication schemes have their architecture based on graphics in different varieties and compositions. This paper first presents several usable authentication schemes that are available, examining the architecture they are based upon. Next is a critique of the architecture of the described usable authentication schemes. Finally, the identified requirement for the described usable authentication systems is reflected upon.

Keywords- Usable authentication, Security, Graphic based password.

1. INTRODUCTION This feature is intended to reduce the memory load on users and is an easier memory recall task than pure recall. In recognition-based It is no doubt that user authentication has become a frequently graphical password systems, users typically memorize a portfolio debated topic in information security as the use of automated of images during password creation and then must recognize their systems has spread its tentacles to almost every path of our images from among decoys to log in. activities. User authentication is one of the important topics in information security to protect users’ privacy [6]. Several In [28] it was opined that based on the assumption that pictures are authentication schemes exist either as text based or graphic based easier to remember than words and a picture is worth a thousand schemes while other authentication models include biometric passwords, many researchers are focusing on the development of scheme, smart cards etc. Usable authentication encompasses all the graphical authentication. All these effort by researchers are in graphical password authentication scheme while graphical order to provide usable authentication schemes. password as defined by [14] is a secret that a human user inputs to a computer with the aid of the computer’s graphical input (e.g., 2. USABLE AUTHENTICATION SCHEMES mouse, stylus, or touch screen) and output devices. A definition of graphical password from whatis.com says it is an authentication Graphical password schemes exist in different varieties and of system that works by having the user select from images, in a different architecture. In a paper titled “Value of visualization” by specific order, presented in a graphical user interface (GUI). [19], it was established that visualization is of great importance in information representation. Graphical passwords which are In a paper written by [28], authentication methods were divided alternatives to text-based authentication scheme with the aim of into three categories which include; combating the incumbent problems to text based scheme as well as  Token based authentication to aid user’s password memorability, represent the usable  Biometric based authentication authentication schemes. The architectures of the existing graphical  Knowledge based authentication password schemes are analyzed as below.

Knowledge based authentication is the most used authentication 2.1 Draw-A-Secret (DAS) method under which graphical user authentication and text-based password fall. With knowledge based authentication, users are The draw-a-secret password scheme as proposed by [18] is a expected to recall a password or identify some set of pass-images. knowledge based authentication that demands that users should be In a broader sense, knowledge based authentication is divided into able to draw their picture password on a 2-D grid with the aid of a two main categories: recall based authentication and recognition stylus or other pointing devices such as mouse. The path on the based authentication. The recall based authentication is further grid which the user follows when registering the picture as divided into pure recall and cued recall. password, must be followed when redrawing at login. With this graphical password scheme, users no longer have to remember Graphical passwords requiring pure recall are most similar to text strings of text and numbers thereby improving on the usability of passwords because users must remember their password and the system. For this reason, draw-a-secret password can be reproduce it without any cues from the system. In cued-recall regarded as a usable authentication scheme. systems, the system provides a cue to help trigger the user's memory of the password (or portion thereof).

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

In [22] Nali and Thorpe, carried out an experiment asking participants to draw 6 “doodles” and 6 “logos” on a 6 x 6 grid. The drawings which were inspected visually for symmetry revealed that participants tended to draw symmetric images with few pen strokes (1-3) and tended to place their drawing approximately in the centre of the grid. A sample draw-a-secret password scheme is shown below.

Fig. 2: Login screen for Pass-Go (Adapted from Tao H. and Adams C, 2008)

2.3 Inkblot authentication This authentication scheme was proposed by Adam Stubblefield and Dan Simon in [2]. The inkblot authentication is not a full graphical password authentication scheme as it uses images/graphics as cues for users to enter strings of text password. At password creation stage, users type in the first and last letter of the word/phrase that best describes the inkblot from a series of Fig. 1: Sample Draw-A-Secret password inkblots shown to them. In inkblot authentication, an algorithm is (Adapted from Jermyn, Mayer, Monrose, Reiter, and Rubin, 1999) used to generate inkblots. This algorithm takes as input, user information such as username as well as information about the 2.2 Pass-Go authentication target (for instance, the server’s DNS name), and a The pass-go authentication scheme was designed by Hai Tao from random seed. The system generates and stores a new set of random the University of Ottawa in Canada. The pass-go scheme is very seeds each time the user changes password, one for each inkblot. similar to the draw-a-secret scheme except that it demands that The inkblot graphical password authentication scheme is user should draw their picture password on the intersections of the illustrated in the following figure. grid and not the in the grids as with draw-a-secret scheme. In addition to the strengths of draw-a-secret scheme, the pass-go scheme makes use of colour pen (to increase the variability of passwords), finer grid and supports diagonal movement unlike draw-a-secret that allows for only vertical and horizontal movement. Users selected longer passwords and used colour, both resulting in greater password complexity than in DAS which in effect renders dictionary attacks less effective than DAS. The following figure shows the interface for the pass-go graphical password authentication scheme.

Fig. 3: Inkblots used in the Inkblot Authentication user study (Adapted from Stubbleffeld A. and Simon D., 2004)

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

2.4 Blonder’s graphical password This authentication scheme was patented to Greg Blonder in [16]. This scheme required users to touch predetermined areas of an image in sequence for authentication. The notion behind the use of an image is for it to act as a cue to help users remember their password. For instance, in the figure shown below, a password could consist of clicking on the camera, clock, the picture on the wall, the money on the bed, and left handle of the wardrobe. The sequence for which it was registered must be followed by the user at login. In [28] it was revealed that, Passlogix had a similar implementation of blonder’s graphical password as a part of their v-GO system which no longer seems to be available. The figure below shows Passlogix implementation of blonder’s password scheme.

Fig. 5: Screenshot of the Deja Vu graphical password system (Adapted from Dhamija. and Perrig, 2000)

2.6 Passpoints Authentication This is a cued recall based authentication scheme. It is a slight variation to the blonder graphical password. At password creation time, users are presented with an image and all they have to do is to select a sequence of any 5 click-points (pixels locations) on it by clicking. During login, re-entry of the click-points must follow the same sequence for which they were made at creation time.

A specified tolerance level is specified for each click-point since it is not possible for humans to click on exactly the same pixel at different times. The image has no other purpose other than to act as aid for users to remember the click points. The implementation

Fig. 4: Passlogix implementation of Blonder's of PassPoints requires that for each click-point, an imaginary grid graphical passwords. is overlaid onto the image; if a guessed click-point falls within the same grid square as the original point, then the guess is accepted. (Adapted from Suo X., Zhu Y., and Owen G., 2005) The number of entries in the theoretical password space for

PassPoints is based on the number of squares in this grid and the 2.5 Deja Vu authentication number of click-points in a password as proposed by [8].A This scheme was proposed by [12]. The scheme is based on the snapshot of the passpoint graphical authentication scheme as use of random arts as graphics for which user will authenticate. adapted from “Modeling user choice in the passpoints graphical The selection of these random arts include very attractive images password scheme” by Ahmet E. et al, is shown in the following so as to increase the likelihood that images have similar figure. probabilities of being selected by users. The architecture of Deja vu as adapted from Dhamija and Perrrig’s “Deja Vu: A User Study Using Images for Authentication” in 2000, is as follows.

Using Deja Vu, the user creates an image portfolio, by selecting a subset of p images out of a set of sample images. To authenticate the user, the system presents a challenge set, consisting of n images. This challenge contains m images out of the portfolio. We call the remaining n - m images decoy images. To authenticate, the user must correctly identify the images which are part of her portfolio (Dhamija and Perrig, 2000). The Deja vu scheme is a recognition based scheme and as such demands that users are able to identify their set of images required for authentication. The Deja vu graphical password scheme is shown below.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Fig.8: Sample panels for the PassFaces graphical password. On theleft is a sample panel from the original system

(Adapted from Davis, Monrose, and Reiter, 2004). On the right, a Fig. 6: Screenshot of the Passpoint scheme panel with decoys similar to the image from the user's portfolio (Adapted from Ahmet E, Nasir M and Jean-Camille B.,2007) (adapted from Dunphy, Nicholson, and Olivier, 2008).

2.7 Passface authentication The passface authentication is a recognition based scheme, 2.8 Jetafida Scheme This authentication scheme, designed by [5], is a recognition based introduced by Passfaces (formerly known as Real User scheme. The algorithm puts together the entire usability feature as Corporation). It is an information security technology company stated by the ISO usability standards; that is, ease of use, ease of based in Annapolis, Maryland. Users are asked to pick their creation, ease of memorizing, ease of learning and acceptable assigned Passfaces from a 3 x 3 grids containing one Passface and design and layout. 8 decoys. The faces appear in random positions within the grid each time. This process is repeated until each of the assigned Consequent to a Survey on Recognition-Based Graphical User Passfaces is identified. (Passfaces website). The operational flow Authentication Algorithms carried out by [15], which includes, of the passface graphical password authentication scheme is shown DejaVu scheme, Jetafida Scheme, Passpoint Scheme, triangle in figure vii. ([20]). A screen capture of passface graphical scheme and several others, the Jetafida scheme which is most password authentication scheme is shown in fig. 8. recent presents to be the most promising in terms satisfying the ISO standards for usability . The Jetafida scheme presents users with a welcome interface that allows existing users to login as well as lets new user to create an account on the system. The interface consists of a grid of twenty images of different variety for the user to select and arrange. The users will select three pictures as a password and then sort them according to the way he wanted to see them in login phase. A screenshot of the login interface of the jetafida authentication scheme is shown below.

Fig. 7: Operaional flow of the Passface scheme (adapted from Kok-Chie Daniel Pu, 2006)

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

3.3 Blonder and PassPoint These authentication schemes use purely graphics and users must click on certain points in a sequence. These approaches to authentication are still very much prone to attacks. Programs that take note of coordinates of click-points on the screen can be used to track the locations that a user clicks during login. Shoulder surfing is another challenge that these graphical password schemes (PassPoint and Blonder’s) are non-resistant to. The entire user password can be viewed on the screen and as such, at one login by the user, every points clicked can be captured by shoulder surfing. Hotspots in the passpoint scheme are said to be point on the graphic/image that has very high probability of been chosen or clicked by users as passpoints. PassPoints scheme is vulnerable to hotspots and patterns within images [25]. Hotspot is a serious security problem in click-based schemes [17]. In a study carried out by [13], it was revealed that users can adequately describe their password (the points to click) to someone else in order to grant him access. The point here is that, both passpoint and blonders graphical schemes are susceptible to social engineering attacks. Fig. 9: Screenshot of the login interface of jetafida scheme (Adapted from: “Graphical password: prototype usability survey, 3.4 Deja Vu Ali Mohamed, E., Norafida. Ithnin.,” 2008 International This system is a full recognition based authentication scheme and Conference on Advanced Computer Theory and Engineering, several studies have shown that humans have outstanding ability to 2008) identify previously seen images even if it was seen for a short time [26], [23]. In [7], it was opined that from a security perspective, 3. CRITIQUES recognition based systems (which include Deja Vu) are not suitable replacements for text password schemes, as they have 3.1 Draw-a-Secret and Pass-go password spaces comparable in cardinality to only 4 or 5 digit Both the draw-a-secret and pass-go authentication schemes lack PINs (assuming a set of images whose cardinality remains resistance to shoulder surfing as the picture being drawn will have reasonable, with respect to usability). This system, though less, is to take some time before completion and such amount of time may non-resistant to shoulder surfing as the whole random arts in the be sufficient for a person to carefully study what is being drawn by user portfolio required for authentication can be viewed on the shoulder surfing. screen after several logins have been observed to identify all of the images in a user's portfolio. In [25] a study showed that a large number of passwords from the experiment involving 16 participants carried out by [22], falls 3.5 Passface and Jetafida within predictable categories and can sufficiently help attackers Peculiar to most recognition based scheme is the problem of identify candidate passwords with high probability of success and shoulder surfing which does not leave out both Passface and launch efficient dictionary attacks. In the same study, the same Jetafida authentication scheme. Jetafida scheme, being a recent findings were made for Tao’s pass-go authentication scheme. graphical authentication scheme is yet to be studied by many Furthermore, “easy to execute” being one of the attributes of researchers. Judging from its architecture it is still not completely “satisfaction” as a usability feature specified by the ISO ([4]), is shoulder surfing resistant. According to a research by [1], the items lacking in the draw-a-secret scheme and can be said to lure users of usability are easy to use, easy to create, easy to memorize and into circumventing the system as users will have to draw each time easy to learn. This research shows that the Jetafida scheme is not they need to login. This, in effect jeopardises the usability of the meaningful and not effective. Passfaces on the other hand, is very system. susceptive to social engineering attacks as shown by [1].

3.2 Inkblot authentication 4. LESSONS AND RECOMMENDATIONS First, the inkblot authentication scheme is not a fully graphical password scheme and as such text strings are involved. The Our experience with the above studied usable authentication presence of text in the password is a great loophole as the schemes has led to a number of realizations that may help future weaknesses of text based password will still apply to it. Another designers of usable authentication systems. It is recommended that thing is that, an attacker, when presented with the same set of a system that seeks to improve users’ trust in the authentication inkblot as an authorised user, may provide the same first and last system by improving the text based authentication with encryption letter of what he thinks the inkblot represents as the authorised and hence aid user memorability by adding graphics making it an user. hybrid system should be developed.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The suggested hybrid system will be able to curb shoulder surfing REFERENCES problems. The combination of encrypted text and graphics makes it a better system. The text password will be encrypted with a one- [1] M. Eljetlawi and N. Ithnin. “Graphical password: Prototype way hash function and stored in a database different from the one usability survey”. In Proceedings IEEE International that hosts the graphics. At each login, the positions on all the Conference on Advanced Computer Theory and Engineering, images will change just like passpoint scheme. The suggested pp. 351-355, 2008. hybrid system should be based on the following principle. It [2] Adam Stubblefield and Daniel R. Simon, "Inkblot should demand three categories of information from the user, Authentication". Technical Report MSR-TR-2004-85, which include: username, a text based PIN and a set of images. August, 2004. The username is required to identify a user, that is, who wants to [3] Ahmet Emir Dirik, Nasir Memon and Jean-Camille Birget. gain access to a particular system or resource. The other two “Modeling user choice in the passpoints graphical password categories which include a PIN and a set of images are used to scheme,” Symposium on Usable Privacy and Security, authenticate the user. The PIN (personal identification number) is Pittsburgh, Pennsylvania: USA, July 2007. system generated and is of four characters in length. The set of [4] Alain Abran, Adel Khelifi, Witold Suryn, and Ahmed Seffah, images contains three user-owned images which must be supplied “Consolidating the ISO Usability Models”. 2003. at login period. [5] Ali Mohamed Eljetlawi, Norafida Ithnin. “Graphical password: comprehensive study of the usability features of The first thing the user has to specify is his username followed by the recognition base graphical password methods,” Third the four-digit PIN which will not be visible on the screen (that is, 2008 International Conference on Convergence and Hybrid no character will be displayed at all) and then he is presented with Information Technology. 1137-1143, 2008. a grid containing about thirty images from which he is expected to [6] Arash Habibi Lashkari and Samaneh Farmand, "A survey on choose three (the ones chosen at registration time) in any order. If usability and security features in graphical user authentication all these are supplied correctly, then the user will be authenticated algorithms", 2009. and allowed access to the system or resource. With this technique, [7] Biddle R, Chiasson S, and van Oorschot P. "Graphical shoulder surfing attempts will not be productive as an attacker will Passwords: Learning from the First Generation " TR-09-09: have to surf both the screen as well as the keyboard. This GRAPHICAL PASSWORDS, 2009. technique will also withstand attacks by software that keep track of [8] Birget J., D. Hong, and N. Memon. Graphical passwords click points as the images will be randomised at every login hence based on robust discretization. IEEE Transactions on they cannot keep track of the encrypted system generated PIN’s. Information Forensics and Security, 1(3):395{399, 2006. [9] D. Davis, F. Monrose, M. Reiter, "On user choice in graphical 5. CONCLUSION password schemes", 13th Usenix Security Symposium, 2004, 1-14. Almost all of the existing usable authentication schemes have one [10] D. Nelson, V. Reed, and J. Walling, “Pictorial Superiority or more noteworthy security shortcomings for which shoulder Effect,” Journal of Experimental Psychology: Human surfing is a major player. The schemes do not provide mechanisms Learning and Memory, vol. 2, No. 5, 1976, pp. 523–528. that will, to a large extent, protect against shoulder surfing. Some [11] Definition of graphical password from whatis.com of these schemes do not suggest good usability as they are difficult http://searchsecurity.techtarget.com/sDefinition/0,,sid14_gci1 for the users to memorize and adapt to. 001829,00.html, accessed Dec 2010 [12] Dhamija, R. and Perrig, A. “Deja Vu: A User Study Using An important goal of all usable authentication schemes is to ensure Images for Authentication”. In Proceedings of the 9th a usable yet secure system for user authentication. Sometimes this USENIX Security Symposium, 2000, 45–48. was done too faithfully, compromising parts of the system aspects of usability; that is, ease of use, ease of adapting, ease of [13] Dunphy P, Nicholson J, and Olivier P. Securing Passfaces for memorizing and several others. description. In 4th Symposium on Usable Privacy and Security (SOUPS), July 2008. To ensure that the goals of usable authentication schemes are [14] F. Monrose, M. Reiter. "Graphcal password", August 2005. achieved, developers should put into full consideration, the users [15] Farnaz Towhidi and Maslin Masrom. "A Survey on whilst ensuring high level security of user’s authentication details. Recognition-Based Graphical User Authentication Algorithms". (IJCSIS) International Journal of Computer Science and Information Security, 2009, Vol. 6, No. 2. [16] Greg E. Blonder. U.S. Patent No. 5559961 [17] Haichang Gao, Zhongjie Ren, Xiuling Chang, Xiyang Liu and Uwe Aickelin. "The Effect of Baroque Music on the PassPoints Graphical Password", 1996. [18] Jeremyn, A. Mayer, F. Monrose, M.K. Reiter, A.D.Rubin, The design and analysis of graphical passwords, Proc. 8th Usenix Security Symposium, 1999.

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[19] Jarke J. van Wijk. "Value of visualization" Technische Universiteit Eindhoven, 2004. [20] Kok-Chie Daniel Pu. Lecture notes on "Graphical Mr. LAWAL Olawale N. is a passwords". April 2006. Available at lecturer in the Department of http://www.cups.cs.cmu.edu/courses/ups- Computer Technology, Yaba College sp06/slides/060413.ppt, accessed Dec 2010 of Technology, Lagos, Nigeria. He [21] L. Standing, J. Conezio, and R. Haber, “Perception and obtained MSc (Computer Science) memory for pictures: Single-trial learning of 2500 visual from University of Lagos, Lagos, stimuli,” Psychonomic Science, vol. 19, no. 2, p. 7374, 1970. Nigeria. His research interest include: [22] Nali D and Thorpe J. May 2004. Analyzing user choice in Genetic Algorithms, Genetic graphical passwords. Technical report, TR-04-01, School of Programming, Software Engineering, Computer Science, Carleton University. Systems Security. [23] Nelson D, Reed V., and Walling J. 1976. Pictorial Superiority Eject. Journal of Experimental Psychology: Human Learning and Memory, 2(5):523{528. [24] Passfaces Corporation, “The science behind Passfaces,” White paper, http://www.passfaces.com/enterprise/resources/white_papers. htm, accessed Dec 2010. [25] Salehi-Abari A., Thorpe J., and van Oorschot P. 2008. On purely automated attacks and click-based graphical passwords. In 24th Annual Computer Security Applications Conference (ACSAC). [26] Standing L., Conezio J., and Haber R. 1970. Perception and memory for pictures: Single-trial learning of 2500 visual stimuli. Psychonomic Science, 19(2):7374. [27] Tao H. and Adams C. 2008. Pass-Go: A proposal to improve the usability of graphical passwords. International Journal of Network Security, 7(2):273{292. [28] Xiaoyuan Suo, Ying Zhu, G. Scott. Owen, "Graphical Passwords: A Survey," acsac, pp.463-472, 21st Annual Computer Security Applications Conference (ACSAC'05)

AUTHOR’S BRIEF

Dr. Olanrewaju O. Ayannuga Biography Olanrewaju O Ayannuga is a Doctor of Computer Science at the renowned Yaba College of Technolgy, Lagos, where he presently is the head of the computer science department. His work focuses in the area of security and usability of computer/software systems. In recent years, his work has included implementation of graphi-text authentication schemes in electronic voting systems. Lately, his works has overlaped between cryptosystems as well as secure authentication.

Research Areas Usability of Security System Secure Usability Design System for User Authentication Usable Hybrid Authentication System Graphical Authentication Mechanism Electronic Voting Systems Cryptosystems Network Security

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Evaluating the Impact of IT Empowerment Initiatives on Income Generation, Employment and Productivity

M.N. Agu Department of Computer Science University of Nigeria Nsukka, Nigeria. [email protected]

ABSTRACT Information Technology is a tool that can help individuals expand their consciousness and capacity for empowering themselves. It provides people with a better understanding of the technology and its usefulness in our every day life. Its use should increase access to information which should lead to possible ways of improving people’s wellbeing. This study is an attempt to see how our rural and urban poor can be empowered, through information technology (IT) to reduce their poverty levels. The study is focused on the impact of IT on income generation, employment and productivity. The research has been conducted using structured questionnaires to get the views of the stakeholders. The responses from the questionnaires are analyzed and interpreted. The findings from the result reveal the following: The poor can be empowered through IT. Empowering people or the youth through IT will increase their income generation. It is observed that most organizations needed people who are IT empowered. IT empowerment is found to increase job opportunities for people and also increase production.

Keywords- Impact Evaluation, Information Technology, Empowerment, Income generation, Employment, Productivity.

1. INTRODUCTION 2. RELATED LITERATURE

One of the greatest and most uncomfortable problems facing the Information Technology has been shown to promote Nigerian economy in recent times is the steady increase in youth economic growth. Since the early 1990’s, information unemployment. According to [12] unemployment in Nigeria is technologies (ITs) and the related services are believed to assuming a crisis-level. Based on the number of graduates and have the potential to promote steady and sustainable secondary school leavers without jobs, it is evident that growth, to increase competitiveness, to open new job unemployment rate is growing at geometric progression [12]. This opportunities and to improve the quality of life to all situation is most disturbing, and when one considers the thousands Europeans [European of youths that graduate from our tertiary institutions each year one wonders whether the Federal Government has really appraised Commission, 1993 in [9]]. [8] discussed how IT provides itself of the serious problems (e.g. armed robbery, drug trafficking unparallel opportunities to improve the lives of the etc) of youth unemployment in this country. Apart from youths world’s poorest people by creating jobs, improving access engaging in various vices such as armed robbery, drug trafficking to health care, providing education and other services. IT and the like, there is loss of social peace, orderliness, security and has helped in connecting community-based artisan other vices associated with poverty. This has challenged us to producer groups[3] . It can create job, improve access to determine the extent IT initiatives can contribute to income health care, education and other services [8]. Village pay generation, employment and productivity which will lead to phones provided a means of communication and a means reduction and empowerment of the poor. The impact of IT of income generation ([6], [5] ,[3]. In Ghana studies initiatives will be determined by its impact on employment and revealed that youth participation in education and skills how economic growth as a result of IT transforms itself into the training is inadequate hence the vulnerability in terms of creation of productive and remunarative employment. Determining employment. In view of this situation the information and the impact of IT on income generation, employment and communication technology sector which has the potential productivity has become a crucial task. With the speed of to generate job opportunities for the youth needs to be communications and the pace of unemployment constantly on the promoted. The regulatory body in the sector has to be increase, the government needs to know how technology is strengthened and given the necessary independence to affecting the youth. The government needs to know whether IT provide an enabling environment for private sector systems are helping the youth or hindering them for achieving their participation for the sector to develop and create the much goals. needed jobs for the youth [8].

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[11] in his paper discussed the use of IT in poverty alleviation. He 2.1 Research Questions argues that the impact of IT on the lives of the poor goes beyond 1. Can the use of IT strategy contribute to income income generation. Ultimately, it is important to evaluate the generation ? impact of IT initiatives on income generation, employment and 2. Is it possible to create employment opportunities productivity. The impact of IT on poverty will fundamentally, be by the use of IT? determined by its impact on employment and how economic 3. How far can the application of IT help in growth, as a result of IT, transforms itself into the creation of enhancing agricultural productivity? productive and remunerative employment. We want to determine 4. The general objective of this work is the the extent the adoption of IT in different sectors of the economy exploitation of IT for creating employment, enhances overall growth and productivity and generate linkages improving income generatoin of people and with activities that provide livelihoods to people, in a way that improving agricultural productivity of the positively impacts on their income generation thereby alleviating farmers. poverty. It has often been felt that lack of access to technology in rural and remote areas, and by poor and under-privileged women The hypotheses for the research are set as follows: and minorities in addition to lower levels of literacy contribute a lot to the subsistence of their poor status. H0: IT empowerment cannot contribute to income generation. According to [6] ICT ( Information and Communication Technology) refers to any artifact, technique or knowledge used to H0: IT empowerment cannot help in creating create, store, manage and disseminate information. This includes opportunities for employment. radio, television, video cameras and telephones. IT is defined as the set of activities that facilitate the capturing, storage, processing, H0: IT empowerment cannot help to improve agricultural transmission and display of information by electronic means [16]. productivity. They include telegraph, radio, television, computers, internet services and wireless technologies. 3.0 METHODOLOGY [1] defines ITs by categorizing them based on how they have been in common use and to some extent the technology used for the Data was collected through the following methods: transmission and storage of information. Questionnaires: New IT- this includes computer, satellites, wireless one-to-one Data was collected through the use of questionnaires. The communication (mobile phones). Four characteristics can be used questionnaires were designed and were distributed to three to describe the modern ITs. categories of respondent namely the youths, parents and employers of labour. What guided the questionnaire are - Interactivity: they are now effective two-way communication the following: technology. The flow of information can be interactive. It is a) The objective of the work no longer only received but can be created and offered more b) The research questions easily. c) The hypothesis to be tested. - Permanent availability: it is available 24 hours a day - Global reach: geographic distances hardly matter any more. A simple random sampling was done. Within this random - Reduced cost: for many people cost of communication has sample a stratified random sample was also taken. The gone down to a fraction of previous values and relatively questionnaires were directed to the stakeholders and the cheap. following parameters were considered: a) Computer awareness /Computer usage Old IT –Radio, Television, land line and what we may call ‘really b) Internet usage old IT’ such as newspapers, books and libraries. Today a useful c) Its application areas working definition of IT can be seen as an electronic means of d) Awareness of job opportunities capturing, processing, storing and disseminating information. [11] e) Its effects on agriculture believe that information and knowledge are critical components f) Whether it is necessary for most job and information technologies (ITs) offer the promise of easy g) Its effect on income access to huge amounts of information useful for the poor. The research questions are as follows: Documents Documents showing how IT initiatives were used in other countries were used[13]. Results from other studies show that IT empowerment can help reduce poverty. The purpose of the research is to see whether IT initiatives can be an effective tool for income generation for both urban and rural dwellers in Nigeria..

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

3.1 Data Analysis The questionnaires were analyzed using SPSS (Statistical Package for the Social Sciences). A frequency distribution of the key variables using a bar chart was drawn. This is shown in fig. 1.

Fig.1: Showing the percentages usage of those key variables.

From fig.1 it is observed that out of the total population, Chi Square Analysis (Test of Independence) 97.2% use computers. It is also observed that 86% use it for This analysis was performed to find out how our respondents word processing and 59.2% use Spreadsheet/Excel, 52.5% answered some questions related to the effect of IT use Presentation software. This shows that most people who empowerment on the key variables. To achieve this we first use the computer use it mainly for Word processing, Excel determine our dependent and independent variables. For the and Presentation software. Out of the total population analysis the independent variable will predict a response with interviewed 84.9% use the internet, 59.9% use it for research the dependent variable. The objective of the study is to while 82% use it for checking mails. It is also observed that a determine the effect of IT empowerment on income, creating major consideration for job offers is the position of IT skills. job opportunities. From fig. 2 it can be seen that 96.3% of The same Fig. 1 indicates that 74.3% of the population those with IT training agreed that it has helped to increase possess IT skills for job offer. their income earning capacity

Fig 2: Showing the bar chart of income generation with respect to IT empowerment.

100

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Test of association between IT empowerment and income generation. H0.: IT empowerment cannot contribute to income generation.

In this analysis IT empowerment is the independent variable while income is the dependent variable.

Table 1: Chi-Square Tests for Income Generation and IT training Value df Asymp. Sig. (2-sided) Pearson Chi-Square 30.213 4 .000 Likelihood Ratio 28.626 4 .000 Linear-by-Linear Association 16.647 1 .000 N of Valid Cases 179

From table 1 Chi-Square value (X2 ) =30.213 P< 0.5 Therefore we reject the null hypothesis and conclude at 95% confidence that IT can contribute to income generation. From Fig.3 it can be seen that 78% of those with IT training agreed that it has afforded them more job opportunities .

Fig.3: Showing the bar chart of creating oppurtunities for employment with respect to IT empowerment.

Test of association between IT empowerment and opportunities for employment. H0 : IT cannot create opportunities for employment.

Table 2: Chi-Square Tests for Discovering Job Opportunities and IT Training Value df Asymp. Sig. (2-sided) Pearson Chi-Square 29.405 4 .000 Likelihood Ratio 28.234 4 .000 Linear-by-Linear Association 9.611 1 .002 N of Valid Cases 179

From table 2 Chi-Square value (X2 ) =29.405 P< 0.1

Therefore we reject the null hypothesis and conclude at 99% confidence that IT training can create opportunities for employment. From Fig.4 it can be seen that 94.8% of those with IT training agreed that it has afforded them opportunities to improve their agricultural productivity

101

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Fig.4: Showing the bar chart of improvement in Agricultral productivity with respect to IT empowerment

Test of association between IT empowerment and improvement in Agricultural productivity.

H0. IT can help to improve Agricultural productivity.

Table 3: Association between Improvement in Agricultural Productivity and IT Training Value df Asymp. Sig. (2-sided) Pearson Chi-Square 19.722 4 .001 Likelihood Ratio 18.331 4 .001 Linear-by-Linear Association 10.186 1 .001 N of Valid Cases 179

From table 3 Chi-Square value (X2 ) =19.722 P< 0.5 To illustrate how useful IT can be for farmers, consider the Therefore we reject the null hypothesis and conclude case of farmers in India who in the past were harvesting at 99% confidence that IT can Improve Agricultural their tomatoes at the same time, giving rise to a market glut productivity. that pushed prices to rock bottom. At other times, when tomatoes weren’t available and the prices shot up, the 4. RESULT OF ANALYSIS farmers had none to sell.[31].Now, they use a network of Telecentres to coordinate their planting so that there is a It is observed that many people use computers. Most of the steady supply to the markets and more regulated and people that use computers seem to concentrate on word regular prices. processing packages. Most organization use word processing in their work, IT training became necessary for IT and Employment Opportunities and Income most employment opportunities. Empowering people with generation IT reduces the search time for jobs. There is a significant Poor people in rural localities lack opportunities for relationship between IT training and income generation. employment because they often do not have access to information. One use of IT is to provide on-line services for 5. CONCLUSION job placement through electronic labour exchanges in public employment service or other placement agencies. In IT and Agriculture productivity this way, unemployed people can use IT to discover job Achieving high agricultural productivity is done by opportunities. In addition, they can also become employed delivering useful information to farmers in the form of crop in the new jobs that are created through the deployment of care and animal husbandry, fertilizer and feedstock inputs, IT. IT is a tool that can help individuals expand their drought mitigation, pest control, irrigation, weather consciousness and capacity for empowering themselves. It forecasting, seed sourcing and market prices [11]. Other provides people with a better understanding of the uses of IT can enable farmers to participate in advocacy technology and its usefulness in our every day life. Its use and cooperative activities. increases people’s access to information which should lead to possible ways of improving their wellbeing.

102

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The bulk of poor people live in the rural areas and [7] Golap Monir (2004). ICT: The best tools for poverty information is packaged for them because they have IT to reduction. (On-Line: help them. This can be achieved through equipping their http://nation.Ittefaq.com/artman/exec/view.cgi/43 community/village hall with television that impacts /9065) knowledge on the poor. IT makes education more attractive [8]Grameen Foundation(2001) Empowering People, using video to teach them in their village halls and this will Changing Lives, Innovating for the World’s Poor. reduce mass literacy. Agro based communities can get (On-line http://www.grameenfoundation.org ) information through their village halls on crops, fertilizers [9] Hannes, S(2000) Skills Shortage vs. Job Creation: A through IT which will help improve on their agricultural Review of empirical Evidence on the Issue of production. IT has an important role to play in reducing ICTs and Employment. poverty by improving the flow of information and (On-line: communication. It is a valuable tool for information sharing http://www.diw.de/deutsch/produkte/publikation/ and awareness raising within the wider community vierteljahrshefte/papers/v_00_4_3.pdf) development to combat poverty and advance international [10] Imoro Braimah and Rudith S. King(2006) Reducing development goals. the vulnerability of the youth in terms of employment in Ghana through the ICT sector. REFERENCES International journal of Education and [1] Alan, G (2005). ICTs for Poverty Alleviation: Basic Development using Information and Tool and enabling Sector. Sida, ICT For Commuincation Technology (IJEDICT),2006, Development Secretariat (On-line: 2(3),23-32 http://www.eldis.org/fulltext/sidiacpoverty.pdf.) [11] Kenny, C (2001) Information and Communication [2] Bhatnagar, S(2000). Empowering Dairy Farmers Technologies and Poverty through a Dairy Information & Services Kiosks, Online:http://www.technowlogia.org/TKL_active Washington DC. _pages2/CurrentArticles/t-right.asp?IssueNumb... [3] Cecchini, S and Shah T (2002) Information and [12] Oyemomi, E.O (2003) An Assessment of Poverty Communications Technology As a Tool for Reduction Strategies in Nigeria 1983 – 2003. An Empowerment. World Bank Empowerment unpublished Ph.D dissertation. St Clements Sourcebook :Tools and practices 1 University. 54p [4] Cecchini, S and Scott, C (2003), Can Information and [13] Roger, H. (2002). ICT for Poverty Alleviation communications Technology Applications Framework. ( Prepared for the Workshop for Contribute to Poverty Reduction? Lessons from UNDP Country Office ICT Programme Rural India, Information Technology for Officers/Focal Points in Asia-Pacific) (On- Development, 10, 73-84. line:http://rogharris.org/ICT for Poverty [5] Digital Opportunity Initiative (2001), Creating a Alleviation Framework. PDF). Development Dynamic: Final Report of the [14] Roger, H. (2004). Information and Communication Digital Opportunity Initiative, 2001 (On-line: Technologies for Poverty Alleviation UNDP – http://www.opt-int.org/framework.html) APDIP(2004) [6] Gerster, R. and Zimmermann, S. (2003) Information [15] United Nations(2005) ESCWA: Information and and Communication Technologies( ICTs) for Communication Technologies for Employment Poverty Reduction? Discussion Paper. (On-line: Creation and Poverty Alleviation in selected http://www.gersterconsulting.ch/docs/ICT_For_P ESCWA member Countries. overty_Reduction pdf). [16] World Bank (2002). Empowerment and poverty Reduction: A source book, Washington DC

Author’s Brief

Dr(Mrs) Monica N. Agu is of Department of Computer Science, University of Nigeria, Nsukka, in the faculty of Physical Sciences. Her research has focused on using Information and Communication Technology on Poverty Alleviation and Modelling the performance of Computer Systems. She can be reached by phone on +2348039329480 and through E-mail [email protected]

103

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

An Online Essay- Based Examination Assessment Model Using Double Blind Marking Technique

1C.O. Akanbi & 2Adetunji, A.B. 1Department of ICT ,Osun State University, Osogbo, Nigeria 2Department of Computer and Engineering, Ladoke Akintola University of Tech., Ogbomosho, Nigeria [email protected], [email protected]

ABSTRACT Most existing online examinations are limited to atomic, closed form assessment units, such as Multiple Choice Questions (MCQ). This allows students with poor knowledge of course concepts to pass examination simply by guessing. This type of examination is unsuitable for testing student true cognitive knowledge. This limitation along with other factors makes MCQ online examination type ineffective in student assessment. To provide a solution to the problems associated with MCQ, an online essay based assessment system is proposed in this paper. This discursive examination system uses double blind marking techniques to assess student performances in examination. The Online Examination architecture was developed and the model formulated was implemented using PHP, HTML and MySQL software. The software developed was implemented on Osun state University Intranet.

Keywords- E-learning, e-asssesment, essay-based examination, double blind marking.

1. INTRODUCTION Assessment systems normally require students showing full spectrum of competencies by reading the answer carefully The advent of web applications into the computing and looking for specific features, such as clarity, logic, and technology has brought about a significant revolution in our key points among others. Often, the best assessment is social life including the traditional system of education and achieved by awarding scores according to explicit ordered examination. Many institutions are beginning to reevaluate categories which reflect an increasing quality of response. their traditional methods and have considered providing These questions should be scored according to a uniform pedagogical materials through the Internet. Web-based grading rubric for greater consistency and reliability. To testing and assessment systems offer greater flexibility than provide a suitable solution to the problems associated with the traditional approach because test could be offered at MCQ, a suitable online essay based assessment system is different times by students and in different locations. As proposed in this paper. This discursive examination system online teaching and learning become widespread, there is a uses double blind marking techniques to assess the growing need for educators to consider modes of assessment performances in examination. [1], [2]. 2. RELATED RESEARCH According to [3], the benefits of online assessment include student motivation, immediate feedback, assess to larger [5] describes the rapid growth of computer technology use in classes effectively [1]. Beyond this advantages, online workplaces and education as inexorable. This technology examination in most tertiary institutions have some offers the potential to broaden educational assessment beyond constraints in terms of quality,validity,reliability and fairness. what traditional methods allow. [6] found that valid and Multiple-choice tests also offer greater efficiency and reliable data can be gained through online ability assessment reliability in scoring than an essay. The major disadvantage when comparing online and paper-based intelligence tests. of a multiple-choice item is that the fixed responses tend to According to [7] context of assessing essays on screen emphasize recall and encourage guessing. In an essay (or demand an enquiry into construct validity; explore whether constructed- response) test, students generate responses that the same constructs or qualitative features of essay have the potential to show originality and a greater depth of performance are being attended to by assessors in different understanding of the topic. The essay also provides a written modes. record for assessing the thought processes of the student. [8] presents an excellent argument for having online courses. However, online exam is limited only to atomic, closed form Their research-based findings support the argument for assessment units, such as Multiple Choice Questions (MCQ). having online courses as well as a detailed analysis of the This makes it possible for a student with poor knowledge of a characteristics of online learners. According to [9], course concept to pass MCQ examination simply by assessment plays different roles in teaching and learning guessing, and is unsuitable in testing students’ true cognitive process. It provides teachers with a means of evaluating the knowledge. This, along with other factors makes MCQ online quality of their instructions. Students also use it to drive and examination type to be ineffective. direct their learning. Online assessments can be offered at different time, location or even different test or different

104

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

students. In many cases, online assessments are carried out numerical marks or identity. E- essay examination model in using an institutional Learning Management System (LMS) this study consist of essay / extended response / short answer such as BlackBoard, WebCT, or an in house product via questions (and possibly a mixture of all three). Also some quizzes, forums and digital assignments[10]. E-assessment questions that require drawings and calculations are also can be justified in a number of ways. It can help avoid the taken care of in this system design. meltdown of current paper-based systems; it can assess valuable life skills; it can be better for users – for example by System Architecture providing on-demand tests with immediate feedback, and To successfully represent the traditional assessment model, perhaps diagnostic feedback, and more accurate results via the following user requirements were put into consideration adaptive testing; it can help improve the technical quality of in the design of the architecture: tests by improving the reliability of scoring. (i) an intuitive and familiar interface for students when The earliest online assessment makes use of Multiple Choice entering extended answers / essays. Question Techniques (MCQ). In the view of [11], MCQ exams can be used not just for testing lower level cognitive (ii) secure, distributed, on-screen marking by multiple skills, but can be implemented to measure deeper markers. understanding if questions are imaginatively constructed. (iii) secure offline marking such that markers can Double blind marking ensure that all the assessments have download data to laptops for marking as and when, been considered thoroughly, conscientiously and objectively, and support the subsequent synchronization of data There are three types of double blind marking according to on upload [2],[12] The System Architecture is shown in figure 1 consisting of (i) Sampled double marking: In this process, all scripts five modules: System Administrator Module, Instructor are firstly marked (in small numbers usually Module, Student Module, Inference Engine and Knowledge by the course leader’ or chief examiner and Base. then a percentage is double marked by a moderator for the purpose of verification. In (i) System Administrator Module best practice the first marker has put the marks This module of the solution controls the entire operation of and comments on the assessed work or the the system. With this module, the administrator can add provided proforma lecturers, instructors and students. The module defines the (ii) Full “seen” or “open” double marking: in this registration process. The system administrator can determine process, all scripts are marked by two markers which kind of questions should be available when setting up but the second examiner marks with a test, period of the exam and other site administrative issues. knowledge of the first marker’s marks and comments. The second examiner is expected (ii) Examiner Module to exercise independent judgment and the final The examiner module handles the setting up of exams. It is marks are awarded by computing the average. only accessible to users with the examiners rights and (iii) Full double blind marking: in this process, all privileges. This module allows the examiner to create new scripts are marked by two examiners and the exams and set up the exam with the required configuration second examiner marks with no knowledge of that fits the exam the user is administering. Examiner can the marks or comments of the first examiner. later re-edit the examination setting to fit into the maybe a This method maximizes independence in new session or curriculum. After setting up the examination, marking. Marks may be agreed by simply the user can then add questions of different types to the averaging the scores from the two independent examination using randomization facility. Also, The markers. examiner set the score profile of each question to be assessed For both open and full blind double marking, when manually and allows setting up a number of examiners to examiners cannot agree, a third party moderator is required. mark the essay type questions. Others include registration of In the past, this role was often given to the ‘external exam candidates, Create/edit/delete candidate groups. examiner’. In this paper, full double blind marking techniques was employed in the assessment. The examiner can also access candidates who have taken a test, the time spent on the test and also their respective scores. The score statistics like average score, pass rate and 3. METHODOLOGY cumulative score can also be viewed by the examiner, setting time limit of the exam and randomizing the questions. When A Full Double Blind Marking Technique (FDBMT) is markers log into Exam Online, they are presented with a list adopted as the assessment model in this paper. This model of questions / papers to be marked. Clicking on a question describes scoring written essay exam without an awareness of brings them to the main marking interface that as shown in other raters and /or visual bias by the result of other graders figure 3.

105

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

STUDENT SYSTEM ADMINISTRAT MODULE

OR MODULE BLIND INFERENCE ENGINE MARKERS MODULE

EXAMINER MODULE

STUDENT DATA STUDENT SCORES BANK QUESTIONS K N O W L E D G E B A S E

Figure 1: System Architecture

(iii) Student Module Inference Engine This module is the aspect that is visible to the ‘exam- This is the intelligent (reasoning) component of the takers’. If the exam is free-for-all, then any visitor to the architecture. The roles which could be broadly categorized web application can partake in the exam. But restricted as coordination of all module components as well as exam will require pre-registration. Examinations to be computation roles. This is achieved through the intelligence taken are visible only to candidates who have the privilege embedded using integrated rule and case based reasoning to take a particular exam. So candidate may be privileged scheme. to take several tests while another may not have any pending test based on their curriculum and department. Exams are available only for a period of time set by the 4. E-ESSAY ASSESSMENT MODEL examiner, and the duration of the examination can also be pre-configured by the examiner. Candidates are served Assessments of the examinations require detailed human questions which can be of different types. marking. This is achieved by posting the answers by a student’s to two independent markers and the average of (iv) Knowledge base. the scores are forwarded to the database. We adopt the following simple model for the generation of the marks for This is the knowledge repository for this application a student. An essay examination conducted for student i is system. It consists of three categories of repositories as marked by marker j and the mark Xi obtained is computed shown in figure 1 which are: as the average of the marks returned by all the markers j . Data Bank  Student /Markers Data The mathematical model is represented in equation (1)  Student Scores n X X  ij (1) i  n j 1 

106

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

When two markers are involved, Marker J = 2 so we particular question, this question is sent to two blind Have: markers and the average of the scores obtained returned by the two markers are stored in the knowledge base by the 2 X inference engine for the final computation. X  ij (2) i  2 j1 7. CONCLUSION

Let the score obtained by the same student i in extended In this paper, the limitation of MCQ examination assessment was highlighted and, online essay examination response / short answer be Pi assessment was proposed as a remedy. The software design was conceptualized; the system was implemented using The total score obtained by the student i is Examscorei: PhP, HTML, and MySQL. A sample essay- based exam was used as a test case (CIT 404) for the implementation. 2 X Examscore  ij  p ( 3) As each student completes a particular question, this i  2 i 1 question is sent to two blind markers and the average of the marks scores returned by the two markers are stored in the if a student has taken continuous assessment m the total knowledge base by the inference engine for the final test score(tts) is computation. The software was implemented on Osun state University Intranet. The response from sample users shows m T that the assessment is better than MCQ. The future tts  ij (4) i  m research work is to develop an intergrated MCQ and essay i1 based double blind marking technique

The final score obtained by the student i is Fscorei: REFERENCES

2 x m T (5) Fscore  ij  p  ij i  2 i  m [1] Ghanashyam Rout1 & Srikanta P. A Case Study on i1 i1 E –Examination in University of odisha

International Journal of Internet Computing 5. CHOICE OF DEVELOPMENT SOFTWARE (IJIC), ISSN No: 2231 – 6965, Volume-1, Issue-

2, 2011 PHP and mySQL were used because of its graphical user [2] A. Fluck, D. Pullen and C. Harper: Case study of a interface (GUI) feature which makes communication with computer based examination system Australasian other user possible by displaying pictures and other Journal of Educational Technology, 25(4), 509- standard objects. It has other enhanced features which 523. 2009. makes it an ideal choice for coding program. Some of its [3] Wales, J. & Baraniuk, R. (2008). Technology opens the features include: doors to global classrooms. The Australian, 2-3 i. It is now object oriented February, p. 27. ii. Application and components written in PHP [4]Bull, J. & McKenna, C. Blueprint for computer- runs on web assisted assessment (London, RoutledgeFalmer , iii. Application developed using PHP run with a 2004. managed runtime environment. Christie, J. Automated essay marking for

content—does it work?, in J. Christie. (Ed.) 7th 6. IMPLEMENTATION International CAA Conference, Loughborough, 8–9 July 2003. The System Model in equation (5) was implemented in this [5] Rowe, K.J. In good hands? The importance of teacher study by storing student data, questions and answers in the quality. Educare News, 149:4-14. 2004. database with the use of mySQL on an Apache server. The [6] Bennett, R. E. Inexorable and inevitable: the continuing students, administrator, examiner interfaces were designed story of technology and assessment. The Journal using HTML, javascript and other appropriate of Technology, Learning, and Assessment. Vol. programming languages; PHP connects the data in the 1(1), 1-24 2002 database with the user interface. The application was [7] Preckel F. and Thiemann H. Online Versus Paper – installed on the university server in the ICT laboratory Pencil of a High Potential Intelligence Test housing about 100 systems connected to the university Saves Journal of Psychology 62(4) 2003. intranet. Figures 2,3 and 4 show the student, administrator [8] Paek, P. Recent Trends in Comparability Studies. PEM and examiner scoring modules. Research Report 05-05, 2005.

A particular essay- based exam was used as a test case (CIT

404) for the implementation. As each student complete a

107

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[9] Allen, I. E. & Seaman, J. Sizing the opportunity: The Authors’ Brief. quality and extent of online education in the United States , 2002 and 2003. The Sloan Caleb.O. Akanbi is a Lecturer in the Consortium, Needham , Massachusetts . Department of Information and Retrieved February 12, 2004 from Communication Technology, Osun http://www.sloan-c.org ,2003 State University, Osogbo, Nigeria . [10] Harvey J and Mogey N(1999) Pragmatic issues when He holds a Ph.D degree in Computer integrating technology into the ssessment of Science from Obafemi Awolowo students. In Brown S, Race, P & Bull J (1999) University, Ile Ife . He is a member of Eds), Computer –assisted assessment in higher Computer Professional Registration education London: Kogan –Page . Council of Nigeria (CPN) .His research interest are e- [11]A.Ricket D. Pullen and C. Harpers Case study of a learning., Agent based System, Mobile Computing. E-mail computer based examination system Austalian [email protected] , +2348033920834) Journal of Educational Technology 25(4) , pg 509-523,2009. .[12] Engelbrecht, J. and HardingA. Combining online and Adetunji Abigail Bola is a Senior paper assessment in a web-based course in Lecturer in the Department of undergraduate mathematics. Journal of Computer science and Engineering, Computers in Mathematics and Science LAUTEC, Ogbomosho. She holds a Teaching, 23, (3), 217-231,2003. Ph.D degree in Computer Science. [13]Pullen D and Cusack BContent Management systems ; She is a member of Computer The potential for open education Fact Sheet, Professional Registration Council of FS01, Australian College of Educators, Canberra Nigeria (CPN). Her research interests are Machine 2008. learning, Data Mining, Artificial Intelligence and Database. She can be reached by phone on +2348034858047and through E-mail: [email protected] [email protected].

108

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

APPENDIX

Figure 4:

Examiner Scoring Page

Figure 2: Student login page

Figure 3 : Student Interface

109

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Support Vector Machine for improving Performance of TCP on Hybrid Network

A. Makolo Department of Computer Science University of Ibadan Ibadan, Nigeria [email protected]

ABSTRACT The Internet transport protocol, Transmission Control Prototol (TCP) by design treats all packet losses as an indication of congestion and reacts to such by reducing its rate. This reduction is not justified on hybrid network where a substantial number of packet losses are due to random errors of wireless link. It leads to underutilization of network resources. Differentiating the cause of packet loss is thus important to enable TCP take actions to control congestion only when the loss is caused by congestion. This work presents the use of a machine learning algorithm-support vector machine (SVM) in differentiating between the two types of losses. The model was built using a labeled dataset consisting of 26,191 loss instances. The SVM model achieved 95.97% accuracy, this shows a substantial improvement in throughput without compromising TCP-friendliness on hybrid network. .

Keywords- TCP, congestion control, machine learning, support vector machine.

1. INTRODUCTION

TCP was originally developed for wired network, where a Explicit loss discrimination can be basically divided into packet loss is an indication of congestion on the network. It Split Connection Approach and Link Level Retransmissions. therefore reacts by reducing its sending rate in order to Split Connection completely hides wireless losses from the control congestion. Wired/Wireless hybrid networks have sender by terminating the TCP connection at the base station characteristics different from wired network. They are prone so that the only losses seen by the sender are congestion to random packet losses that occur not only as a result of losses. It has the advantage of shielding the sender from congestion but other link errors. TCP however has no wireless losses. However, [1] shows that it results in poor end mechanism for differentiating congestion induced losses from to end throughput. Examples of split-connection based error induced losses. It therefore reduces it sending rate each approaches are M-TCP [2] and I-TCP [3]. Link layer time a packet loss occurs. This leads to an under utilization of retransmission hides congestion related losses from the TCP network resources on hybrid network. sender by treating the problem of losses locally. If a packet is To improve the performance of TCP on hybrid network there lost in the wireless link, a local retransmission is performed is need for a mechanism that enables it distinguish between without letting the fixed host know. the two types of losses. In this paper we explore the use of SVM in building a The problem is thus solved at the link layer and the transport classification model for the differentiation of the two types of layer need not be aware. Link-layer retransmission has the losses. advantage that it fits naturally into the layered structure of The remaining part of this paper is organized as follows. network protocols. However, in [4], it is shown that the Related works are described in Sect.2, the classification competing retransmission by the link layer will only lead to algorithm, training and testing dataset in Sect.3.Sect.4 gives significant performance degradation if packet loss rate is the result and analysis. Finally, conclusions are drawn in more than about 10%. Also, there are cases where link level Sect.5 retransmissions are unsuitable. These include interactive real- time applications such as VoIP or video-conferencing. 2. RELATED WORKS Implicit loss discrimination determines the type of packet A lot of effort has been made to differentiate between loss by assumptions and calculations made on measurements congestion induced losses and error induced losses on hybrid of network parameters which can be easily obtained at the network. These can be categorized into two: explicit loss end systems. This method requires the least amount of discrimination and implicit loss discrimination. Explicit loss change in the network and can be implemented by endowing discrimination relies on the support of network apparatus to one of the terminal devices with a classification algorithm. perform differentiation. In this case, the sender is explicitly Loss differentiation decision can be made based on adhoc informed of the type of packet loss by network routers or the rules formed using network parameters such as RTTs, end receiver. ROTTs, and IAT.

110

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

For example, Biaz [5], mBiaz [6], and the statistical packet loss discrimination (SPLD) [7] use packet inter-arrival times, the Zigzag [7] and the Spike [3] use relative one-way trip times, and the Vegas predictor [8] uses round trip times. The limitation of these methods is that the decision-making thresholds are difficult to be determined. Also, studies [9] have shown that using only one feature may not provide a good accuracy.

Another approach is to combine multiple features and apply algorithms from data mining and machine learning. These can learn the relationship between the features and make Figure 1.SVM binary classification decision based on a derived model. Machine learning approach was first used in [10].Multiple features were Given a training set of feature-labeled pairs (xi, yi), i = 1, combined from which a learning algorithm automatically 2,…..l represents the features, xi a loss instance and yi the builds a model. Based on the derived model, congestion loss label denoting the type of loss. yi is either 1 denoting a loss and wireless loss can be differentiated. due to congestion or -1 denoting a loss due to a link error.

3. METHODOLOGY SVM can be used to learn a linear classifier

3.1 Support Vector Machine (1) SVM is employed here to classify the loss causes. Support Vector Machines is an effective statistical learning method for pattern recognition [11]. The SVM based on statistical where w is a weight vector and b is a bias. When the training learning theory has many advantages. One, unlike other samples are not linearly separable, the sample is mapped to a nonparametric techniques such as nearest-neighbors and higher dimensional feature space (Ф)(x): neural network that are based on the minimization of the empirical risk, SVM operates on another induction principle, (2) called structural risk minimization, which can overcome the problem of over fitting and local minimum and gain better before a linear classification is performed in the new space. generalization capability. Two, Kernel function method applied in SVM overcomes the problem of dimensionality effectively without increasing the computational complexity. The problem can be solved as the following optimization Three, SVM has demonstrated higher generalization problem: capabilities in high dimensional space and spare samples. Four, unlike many learning algorithms, SVM leads to good performances without the need to incorporate prior information. Moreover, the use of positive definite kernel in the SVM can be interpreted as an embedding of the input space into a high dimensional feature space where the Subject to ( )+b) 1- ξi classification is carried out without using explicitly the ξ (3) feature space. i

A classification task usually employs training and testing data sets that consist of several data instances. Each instance in the The equation can be rephrased as training set contains one target value (class labels) and several attributes (features). The goal of SVM is to produce a model that can predict the target values of the data instances in the testing set given the attributes only. Support Vector Machines are based on the concept of decision planes that 0 ≤ αi ≤ C, i = 1,…..,l, define decision boundaries. A decision plane is one that subject to yTα = 0 (4) separates between a set of objects having different class memberships, the SVM modeling algorithm finds an optimal where ξi is the slack variable that indicates tolerances of hyper plane with the maximal margin to separate the two misclassification and C>0 is a constant that balances classes. This is illustrated in Fig. 1. maximizing the margin and minimizing the amount of n.e is the vector of all ones, Q is an l x l positive semi-definite matrix,

111

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

For selecting the best parameter value, the grid.py in the Qij ≡ K , and ≡ libsvm-3.12\tools directory was used. ) is the kernel. The decision function is grid.py is a parameter selection tool for C-SVM classification using the RBF (radial basis function) kernel. The training set is divided into five subsets of equal size. Sequentially, one subset is tested using the classifier trained on the remaining four subsets. Thus, each instance of the (5) whole training set is predicted once. The cross-validation accuracy is the percentage of correctly classified data. The 3.2 Description of dataset optimal parameters obtained using the above process are C = 128 and γ = 0.5. These parameters are used to conduct a more The dataset used in this paper is the one generated and used efficient training and to accurately predict unknown data. in [10] .The database was generated by simulations with the After cross validation, the best parameters obtained was used network simulator ns-2 and contains 35,441 loss instances. to train the dataset. The model achieved a cross validation 22,426 of which are due to congestion and the rest to link accuracy is 96.184%. error. The network topologies, the number of wireless links, their place in the topology, the error model and the loss rate 3.4 Performance metric were drawn at random. The following metrics were used for the performance evaluation of the model: The parameters measured at the end of a loss event are the True positive (TP): This refers to the group of positive inter-arrival times and the relative one way delay. The inter- instances that are correctly classified by the algorithm as packet times denotes the arrival time difference between positive. In this case this refers to the number of congestion consecutive received packets. The one-way delay, computed losses that are truly identified as congestion. by one of the two entities, is the difference between the timestamp of the acknowledgement and the timestamp of the True negative (TN): This refers to the group of negative TCP packet, and is actually the real one-way delay minus the instances that are correctly classified by the algorithm as difference between the clocks of the sender and the receiver. negative. In this case this refers to the number of wireless To make the model independent of the absolute values of losses that are correctly classified as wireless losses. these measures, the values were normalized in different ways using the average, the standard deviation, the minimum, and False negative (FN): This refers to the group of positive the maximum of the one-way delay and inter-packet time. instances that are wrongly classified by the algorithm as The motivation for using this dataset, is for the random negative. In this case this refers to the number of congestion generation of the network parameters. This will make the losses that are misclassified as wireless losses. This is model as applicable as possible under different network denoted as ERRC. condition. In addition, it provides a standard for the comparative evaluation of our result with that obtained by False positive (FP): This refers to the group of negative [10]. instances that are wrongly classified by the algorithm as positive. In this case this refers to the number of wireless losses misclassified as congestion losses. This is denoted by 3.3 Training process ERRE LIBSVM 3.12 is used to train and optimize the SVM model. LIBSVM is integrated software for support vector Accuracy: This is the number of correctly classified classification(C-SVC, nu-SVC), regression (epsilon-SVR, instances over total number of instances. As shown in Figure nu-SVR) and distribution estimation (one-class SVM). The 4.1, this is dataset was randomly divided into two; the training set and TP +TN the testing set. The training set contains 26,191 loss instances P+N selected at random. The testing set contains the remaining Misclassification Rate: This is the total number of wrongly 9250 loss instances. Light Data Agent application was used classified instances over total number of instances. to put the dataset into the appropriate format. FP +FN The RBF kernel was chosen. This is the most appropriate P+N here because the dataset is quite large and contains very few number of attributes compared with the number of instances. Also, RBF kernel nonlinearly maps samples into a higher dimensional space and so it can handle the case where the relation between class labels and attributes is nonlinear.

112

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

4. RESULT AND ANALYSIS The SVM model provides an estimate of the probability of each class PC and P E such that PC+ PE = 1 for any given instance of loss. PC is the probability of classifying the loss as congestion and P E is the probability of classifying it as a link error.

The class given by the model is a link error if PE is greater than a threshold PTH and Congestion if otherwise. By default, the value of PTH is fixed to 0.5 so as to treat each class fairly. However, by changing the threshold, one can easily favour the accuracy of the prediction of one class over the other.

Accuracy: The ability of the classifier to correctly identify the category to which each of the loss belongs to given a new situation of a loss event is very important. The higher the Figure 2 .Graph of Accuracy against threshold accuracy the better the performance of TCP + classifier. The SVM model used here achieved an accuracy of 95.97% on the testing data at a threshold of 0.5.This is quite high and means that the TCP+ classifier has a 0.9597 probability of making the right decision when faced with a new loss situation. The accuracy of the classifier is shown in Fig.2 as the threshold is varied.

Misclassification rate: an algorithm that attempts to classify each loss into one of two classes can be judged by its misclassification rate, the fraction of cases which are classified incorrectly. the svm model here misclassifies only 4.03%. however, since misclassifying a wireless loss as a congestion loss does not have the same impact as the other way around, the performance is judged by examining the two misclassification rates separately.

Error on congestion losses (ERRC ):.This is the % of congestion losses misclassified as wireless losses. The effect Figure 3. Graph of error on link error against error of this is that the rate will not be reduced when the network is on congestion congested and can lead to congestion collapse if too high. If there is congestion, the sending rate should be decreased to maintain TCP friendliness. At a threshold of 0.5 the classifier misclassifies only 5.52% of congestion losses. Again this is TCP Friendliness low enough. To maintain TCP-Friendliness, the TCP+classifier should Error on wireless losses (ERRE): This is the % of wireless have a throughput belonging to [1/KBtcp, KBtcp] with K ≤1.78 loss misclassified as congestion loss. Misclassifying wireless [12] where Btcp is the throughput of TCP following the same loss as congestion loss does not cause congestion problems path as the TCP+classifier. Let p be the proportion of the for the network, but it will limit the protocol’s ability to packets loss on a network.The TCP+classifier is a normal improve throughput in the case of wireless network. At a TCP except that it reacts only to a proportion p (1- ERRC) of threshold of 0.5 the model misclassifies 26.2% wireless packets instead of p as in the case of a normal TCP. This losses. means TCP+classifier will only reduce its congestion rate for a fraction 1- ERRc of p. Therefore it can be seen that to Fig.3 shows the graph of the two types of errors when plotted maintain TCP friendliness, the misclassification of against each other with threshold ranging from 0.05 to congestion should not be too high. If the misclassification of 0.95.As seen an increase in one leads to a decrease in the congestion remains close to zero, the classifier will be TCP other The curve is very close to the origin showing lower friendly. values of errors on both sides. This implies a good model.

113

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

However from figure 4.2, it can be seen that the lower the value of ERRC, the higher ERRE. If ERRE goes too high, the gain in terms of wireless link will be compromised. How low should ERRC be without losing TCP Friendliness and at the same time maintaining an optimal throughput on wireless link?

[13] proved that if ERRC is kept below 18%.The TCP+classifier will maintain both TCP friendliness and an optimal throughput on wireless link. The value of ERRC obtained at a threshold of 0.5 is much lower than 18%.

Evaluation of the optimum threshold Fig.4 below shows the change in the ERRC as the threshold is varied. As explained above, to maintain TCP friendliness ERRC should be kept below 18%.This means that the threshold should always be lower than 0.92.

PTH%

Figure 4 .Graph of Accuracy of both losses against threshold

In [10] several machine learning algorithms were trained and tested on the same dataset. It was found that decision tree boosting achieved the highest accuracy of 93.66. SVM has a higher accuracy than this(95.97%).

5. CONCLUSION

SVM is employed to differentiate between the two types of losses that occur on hybrid network. The SVM model Figure 4 .Graph of error on congestion against obtained showed a great capability in classifying correctly a threshold new loss instance. The result showed that when incorporated into TCP it improved the throughput of TCP substantially on Throughput: In order to have an optimal throughput, the wired/wireless networks without compromising TCP- friendliness. correct classification of a loss as due to a link error (Twireless ) must be high. REFERENCES Using the equation proposed by [14]. r is the transmission rate and l is the loss event rate,(the equation is already [6] H. Balakrishnan, V. N. Padmanabhan, S. Seshan, and R. H. Katz, “A comparison of mechanisms for improving defined in 1.2). From this, it can be seen that a TCP performance over wireless links,” Computer misclassification of wireless loss as congestion loss decreases Communication Review, vol. 26, no. 4, pp. 256–269, the throughput by a factor of l . The current TCP Aug. 1996. misclassifies every loss due to a link error and thus reacts to [7] K. Brown and S. Singh, “M-TCP: TCP for mobile all the losses l decreasing the sending rate for each. cellular networks,” ComputerCommunication Review, For a threshold of 0.5, The SVM model correctly classifies an vol. 27, no. 5, pp. 19–43, Oct. 1997 approximate of 73 out of every 100 losses due to a link error. [8] A. Bakre and B. R. Badrinath, “I-TCP: indirect TCP for This is good enough and as seen in Fig.4, the accuracy mobile hosts,” in Proc. 15th Int. Conf. on Distributed increases with the threshold. At a threshold of 0.85 the two Computing Systems, Vancouver, Canada, May 1995, pp. curves intersect. This threshold is lower than the threshold 136–143 that is required to maintain TCP friendliness (0.92) and also [9] A. DeSimone, M.C. Chuah, and O.C. Yue, “Throughput provides a wireless accuracy of 87.15% which is high enough performance of transport-layer protocols over wireless LANs,” in Proc.IEEE Globecom ’93, Nov.1993 for optimum throughput.

114

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[10] Biaz S, Vaidya N H. Discriminating congestion losses from wireless losses using inter-arrival times at the receiver. Proceedings of the IEEE Symposium on Application-specific Systems and Software Engineering and Technology (ASSET’99), Mar 24-27, 1999, Richardson, TX, USA. Piscataway, NJ, USA: IEEE, 1999: 10-17 [11] Cen S, Cosman P C, Voelker G M. End-to-end differentiation of congestion and wireless losses. IEEE/ACM Transactions on Networking, 2003, 11(5): 703-717 [12] Park M K, Sihn K H, Jeong J H. A statistical method of packet loss type discrimination in wired-wireless networks. Proceedings of the 3rd IEEE Consumer Communications and Networking Conference (CCNC’06), Jan 8-10, 2006, Las Vegas, NA, USA. Piscataway, NJ, USA: IEEE, 2006: 458-462

[13] Brakmo L S, O’Malley S W, Peterson L L. TCP vegas: New techniques for congestion detection and avoidance. Proceedings of Conference on Applications, Technologies, Architectures and Protocols for Computer Communication (SIGCOMM’94), Aug 31-Sep 2, 1994, London, UK.New York, NY, USA: ACM, 1994: 24-35 [14] Fu Z H, Greentein B, Meng X Q, et al. Design and implementation of a TCP-friendly transport protocol for ad hoc wireless networks.Proceedings of the 10th IEEE International Conference on Network Protocols, Nov 12- 15, 2002, Paris, France. Piscataway, NJ, USA: IEEE, 2002: 216-225 [15] P. Geurts, I. El Khayat, and G. Leduc. A machine learning approach to improve congestion control over wireless computer networks. In Proc. of IEEE Int. Conf.on Data Mining (ICDM-2004), pages 383–386, 2004 [16] Burges C J C. A tutorial on support vector machines for pattern recognition. Data Mining and Knowledge Discovery, 1998, 2(2): 1-43 [17] S. Floyd, M. Handley, J. Padhye, and J. Widmer. Equation based congestion control for unicast applications. In Proceedings of SIGCOMM 2000, pages 43-56, 2000

[18] El Khayat, P. Geurts, and G. Leduc. Enhancement of TCP over wired/wireless networks with packet loss classifiers inferred by supervised learning. unpublished 2005. [19] J. Padhye, V. Firoiu, D. Towsley, and J. Kurose. Modeling TCP Reno performance: a simple model and its empirical validation. IEEE/ACM Transactions on Networking, 8(2):133-145, 2000

115

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

A Phone Learning Model for Enhancing Productivity of Visually Impaired Civil Servants

Azeta A. Ambrose Covenant University, Ota, Nigeria, [email protected] Phone: +2348039540844

Azeta I. Victor National Productivity Center, Kaduna, Nigeria [email protected], Phone: +2348032613210

ABSTRACT Phone-based learning in civil service is the use of voice technologies to deliver learning and capacity building training services to government employees. The Internet revolution and advancement in Information and Communications Technology (ICT) have given rise to online and remote staff training for the purpose of enhancing workers productivity. The need for civil servants in Nigeria to develop capacity that will enhance knowledge is a key requirement to having competitive advantage in the work place. Existing online learning platforms (such as web-based learning, mobile learning, etc) did not consider the plight of the visually impaired. These platforms provide graphical interfaces that require sight to access. The visually impaired civil servants require auditory access to functionalities that exist in learning management system on the Internet. Thus a gap exist between the able-bodied and visually impaired civil servants on accessibility to e-learning platform. The objective of this paper is to provide a personalized telephone learning model and a prototype application that will enhance the productivity of the visually impaired workers in Government establishments in Nigeria. The model was designed using Unified Modeling Language (UML) diagram. The prototype application was implemented and evaluated. With the proposed model and application, the visually and mobility impaired worker are able to participate in routine staff training and consequently enhances their productivity just like their able-bodied counterparts. The prototype application also serves as an alternative training platform for the able-bodied workers. Future research direction for this study will include biometric authentication of learners accessing the application..

Keywords- Able-bodied, Civil Servants, Phone Learning, Productivity and Visually impaired.

1. INTRODUCTION Phone learning technology is a type of e-learning that is often referred to as voice-enabled learning system. It is the use of land There are various advancement in technology and several e- or mobile phone to access learning content on the Intranet or learning applications such as Moodle, Sagai, Blackborn, etc, are Internet depending on network coverage availability. It uses now available on the Internet. Unfortunately, the visually technologies such as speech recognition and text to speech impaired workers are not among the beneficiaries of these (TTS) conversion to create a user interface that enables users to technologies. Rather, what the visually impaired workers require navigate through a dialogue system using telephone and voice for online learning is assistive technology to provide auditory commands. Phone applications have been developed in several access to the e-learning functionalities on the web. Using the areas such as in e-learning [3,4], banks transactions [5] and a lot web technology as a tool for training development is one of the more. Some other e-learning phone-based applications that have key factors to improve productivity. Investment in human been implemented includes: [6,7,8,9]. Generally, there is dearth development, employee training and technological of phone application in the domain of e-learning and the few modernization are areas that deserve serious attention in existing once lack sufficient intelligence to provide worker’s productivity. This is because, one of the most efficient personalization services beyond learners’ queries and request. and effective methods of improving workers productivity is simply to train them in the skills they need to perform their job Personalized learning is an attribute of independent or duties [1]. There is no doubt that the goal of e-learning in the autonomous learning. Autonomous phone learning platform is workplace is to enhance individual and organizational one of the ways management could use to encourage higher performance [2]. productivity for the visually impaired workers. For instance, as a staff of a government institution, one may be scheduled to attend Enhancing the capabilities of individual staff members is the key training at the Federal government head quarters in the capital to improving overall workforce performance. It is a well - city. The cost and logistics involved in such trip may be saved known fact that low productivity is a national economic when the staff is made to simply dial a telephone number to development problem especially in the public sector. connect to the training materials and receive the lecture through voice interaction.

116

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

When government employees that are visually or mobility etc, but they do not consider the plight of the visually impaired in impaired are not left out in the scheme of training, productivity their interface design and methodological approaches. There are is indeed enhanced to their benefit just like their able-bodied pedagogical issues when group of people engage in e-learning. The counterparts. This research, aimed to provide a personalized knowledge level of the learners differs, hence there is need to Phone learning assistive technology to enhance the productivity deliver learning content to learners based on individual knowledge of the visually impaired employees, and also serve as an level and physical ability to access the learning content. This type of alternative training platform for the able-bodied workers in the e-learning service delivering is often referred to as learning content Nigerian civil service. In workplaces, employees are adults who personalization based on users’ profile. have good self-concept, and usually they have a clear understanding of their learning needs and can learn In a competitive work environment, finding ways to use independently [10]. Through e-learning systems, learners can employee training program to help enterprises adapt to changes study course contents in an independent manner. They can also in the external environment is an important issue [10]. With the decide when to study, the sequence of the content to study, and prospect of cost-effective investment in training, many the amount of time to spend for self education, without time and enterprises have adopted e-learning systems for employee space barriers [11]. Recently, e-learning systems have been training to assist in their human capital management in recent increasingly adopted by organizations for employee training for decades [18]. Through the use of e-learning systems, employees reasons of cost reduction and convenience. can transfer what they have acquired from the training to their jobs and thereby increase their productivity [19]. This helps 2. PRODUCTIVITY AND E-LEARNING IN CIVIL employees’ renewals of knowledge and skills while also SERVICE reducing knowledge gaps between what the organisations have and what the employees need in order to have competitive Technological advancement have given rise to the online staff advantage. training that hold great promise for productivity enhancement [12]. This advancement brought about the need for employees to 3. PROBLEM STATEMENT develop capacity and enhance knowledge for competitive advantage. Globally, an average employer of labour will expect A number of voice-based learning technologies exist such as: the workforce to be productive. Productivity as a concept means [20,4,6,7,9]. Their biggest disadvantage is the rigid structure that different thing to different people. Productivity is defined as the they impose on the end user. While it is convenient to use mobile efficient use of resources such as labor, capital, land, materials, telephony application, it can be extremely slow when the user is energy and information in the production of various goods and forced to drill through several layers of options before finding services [13]. Productivity is the relationship between the exactly what he/she wants [21]. The usage of Phone application is amount of one or more inputs and the amount of outputs from a more rigorous when the visually impaired learner is involved as a clearly defined process [1]. result of their sight impediment. Some level of learning content adaptation using Artificial Intelligence (AI) is required to serve Productivity is thus, of fundamental importance to the individual the purpose of personalization for the visually impaired group of worker of whatever status, to the organization whether commercial learners. This study, thus, provided a personalized voice-based or not, to the national economy at large, to the upliftment of the learning system otherwise known as ‘Phone learning’ to deliver welfare of the citizen and the reduction if not total eradication of learning content through voice response that suits individual mass poverty [14,15]. In addition, it is recognized as a panacea for a visually impaired user, by simply dialing a mobile telephone country's economic woes or ills, it is an attitude of the mind, it is a number. springboard for economic growth, it improves overall standards of living. Indeed, nations of the world now anchor their development 4. THE PHONE LEARNING MODEL policies and planning strategies on productivity and sustainability. Against this backdrop, the concern for productivity improvement The Phone learning model shown in Figure 1 was designed using especially in the public sector has increased with intensity, Unified Modeling Language (UML) deployment diagram. The culminating to the establishment of the National Productivity Centre assumptions of the model are as follows: i) There exist some (NPC) under the Federal Ministry of Employment, Labor and visually impaired civil servants as employees in the Nigerian civil Productivity [16,17]. service, ii) Government employees in Nigeria make use of mobile phones as their means of communication, and iii) Government NPC is a parastatal charged with the statutory mandate to implement employees in Nigeria are required to attend routine training outside productivity improvement program aimed at making productivity their work station. The model shown in Figure 1 has components the driver of our economic activities both in the public and private that require specific description. The visually impaired civil servants sector. The functions of NPC include inter-alia capacity building connects through a mobile phone to the Phone Gateway. The text to and training services to client organizations, both private sectors and speech (TTS) and automatic speech recognition (ASR) executes the civil service. The increasing practice of using e-learning in the civil call interaction with a caller. The Phone Gateway communicates via service has however propelled a gap between two group of web protocols (HTTP) to Phone application server. The web employees - the physically challenged particularly the visually application server queries the database via apache to dynamically impaired and the able-bodied employees. Several technology-based retrieve information and the TTS translate the lecture content from application exist for training such as web-based mobile learning, text to speech for the caller to hear.

117

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Figure 1: A Phone Learning Model for the Visually Impaired Civil Servants

The upper layer (web technology) represents the learning management system (LMS) comprising Moodle, WebCT and Blackboard. The middle layer (AI technology) represents the intelligent personalization of learning content using individual employee profile. The bottom layer (voice technology) comprise of voice interface, phone gateway and the phone application server. The implementation of the AI technology is contained in [20] and it uses case-based reasoning (CBR) and stemming algorithm.

4.1 The Phone Interface Design and Implementation

Authenticated user of the application will undergo some questions and answers session, which will be matched against the content of the database, and the result received by the caller through interactive voice response (IVR). The partial call flow for the phone learning application is shown in Figure 2.

Figure 2: Partial call flow of the application

The model was implemented using the following tools: VoiceXML for the Voice User Interface (VUI), PHP for Web User Interface (WUI), Apache as middle-ware, and MySQL database as back-end. The phone gateway used was provided by [22] for application development and testing. The choice of these tools is due to their advantage as Free and Open Source Software (FOSS) [23]. VoiceXML’s wide acceptance as a standard, huge industry uptake, suitability for multimodal interaction, and increased developer productivity clearly demonstrate its superiority over other tools for developing voice applications [24].

118

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The VoiceXML prototype application was deployed on a Voxeo Several usability studies suggest the system with “Very Bad voice server [25] on the web and accessed from a mobile phone Usability” should have 1 as mean rating, “2 as Bad Usability”, 3 using the format: . Evaluation of the Phone learning system was carried should have a mean rating of 4 on a 1-5 scale and 5.6 on a 1-7 out using [26,27] usability criteria. Statistical Package for scale. Therefore, it can be concluded that the prototype Social Sciences (SPSS) was used to generate the usability application developed to validate our model has “Good analysis and reports. Usability” based on the average (AVG) total rating of 4.03.

5. SYSTEMS EVALUATION 6. RESEARCH IMPLICATIONS

The application was evaluated for usability to determine the The research contain herein has several implications. Lecture level of effectiveness, efficiency, users’ satisfaction, learnability materials are stored in a central server and delivered to learners and memorability. In conducting the usability evaluation, through voice interface. To participate in the training, employees questionnaire were designed and administered. Each section of are required to dial a telephone number from a remote location effectiveness, efficiency, user satisfaction, learnability and using a mobile phone that will connect the caller to the lecture memorability contains five questions represented by Q1, Q2, materials. Expectations for continuous performance of Q3, Q4 and Q5. The questionnaire aims at eliciting information government employees is high, hence the need for an appropriate from learners to measure the usability of the Phone-based training to enhance workers productivity. The aforementioned learning application provided. A total of 70 questionnaires were highlights the need to design training programs that addresses not administered but only 63 responses were received, analyzed and only the needs of the able-bodied learner but also that of the reported. The questions were designed using five-point likert- visually impaired. scale where 1= strongly disagree, 2 = disagree, 3= undecided, 4 = agree and 5= strongly agree. The usability attributes analysis The phone learning system have the following benefits to the of the evaluation has effectiveness 3.96, efficiency 4.12, user able-bodied and the visually impaired: (i) Learners can learn at satisfaction 4.03, learnability 4.00 and memorability 4.02 (see Anytime, Anywhere and Anypace (A3), (ii) productivity is Figure 3). The average total rating of the usability evaluation is enhanced, (iii) the time, cost and logistics of bringing people 4.03. together is eliminated, iv) the new system can reduce training related expenses such as travel, accommodation and facilities, and (v) Government due process policy is avoided.

The issues to consider when using phone learning system are as follows [3]: (i) A new level of competence and awareness with 3.96 4.12 4.03 4.00 4.02 ICT is required for the employee, (ii) diverse opinion on high cost of deployment, (iii) There may also be resistance from government employees as they are used to attending classroom/hall session training outside of their office station, preferably in oversea country.

The opinion of some school of thought about high cost of investing in Phone learning application is arguably incorrect. Research shows that a simple one percent increase in productivity typically produces more than ten times the impact of a one percent decrease in training costs [29]. The requirement is for participating institution to subscribe to a service provider by paying a periodic token of fee.

7. CONCLUSION AND FURTHER WORKS

In this paper, a model and prototype application has been provided to brigbe the gap that exist between the visually impaired worker and their able-bodied conterpart in terms of productivity enhancement. The prototype application was tested Figure 3: Usability attributes analysis by some learners, and the result of evaluation shows that the prototype application developed has “Good Usability” rating of 4.03 out of 5 scale.

119

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The findings shows that the users are enthusiastic about using a REFERENCES mobile phone training tool as another form of assistive technology to compliment the conventional travelling-based [1] Ehiorobo, I. R. (2004), “the implication of training method. When employees of government are made to adequate motivation on workers’ productivity in embrace the new Phone learning system, it follows that the an organisation”, a dissertation submitted to st. efficiency and productivity of staff are enhanced which will clements university in partial fulfillment of the have a multiplier effect on the economy in general. requirements for the award of doctor of philosophy st. Clements University. September The future research direction for this study is to engage voice 2004. biometric techniques based on speech data. Voice authentication will be added as additional means of security [2] Rosenberg, M. J. (2006). Beyond e-learning: mechanism to enhance the authentication of candidates for approaches and technologies to enhance examination and quizzes. An improved system will combine organizational knowledge, learning, and the conventional pin/telephone number authentication and performance, San Francisco: Pfeiffer. biometric to increase security of the application. [3] Azeta, A. A., Ayo, C. K., Atayero, A. A. & Ikhu- ACKNOWLEDGEMENT Omoregbe, N. A. (2009a) “Application of VoiceXML in e-Learning Systems”, Cases on Special thanks go to the Voxeo Inc. for hosting the prototype Successful E-Learning Practices in the version of the application. Developed and Developing World: Methods for the Global Information Economy. Published by IGI Global. Available online at: http://www.igi-lobal.com/reference/details.asp? ID=35211&v=tableOfContents [4] Gallivan. P., Hong, Q., Jordan, L., Li, E., Mathew, G., Mulyani Y., Visokey P. and Tappert C., (2002), “VoiceXML Absentee System, Proceedings of MASPLAS'02. The Mid- Atlantic Student Workshop on Programming Languages and Systems Pace University.

[5] Azeta, A. A., Ikhu-Omoregbe, N. A., Ayo, C. K., & Atayero, A.A. (2008), “Development and Deployment of VoiceXML-Based Banking Applications”, Journal of Computer Science & Its Application, An International Journal of the Nigeria Computer Society (NCS), Vol. 15 No.1 June 2008 Edition, pp. 59-72

[6] Motiwalla, L.F. and Qin J. (2007),” Enhancing Mobile Learning Using Speech Recognition Technologies: A Case Study”. Eighth World Congress on the Management of eBusiness, 2007. (WCMeB 2007). Volume , Issue , 11-13 July 2007 Page(s):18 – 18

[7] Motiwalla L. F. (2009), “A Voice-enabled Interactive Services (VòIS) Architecture for e-learning”, International Journal on Advances in Life Sciences, vol 1 no 4, year 2009 Page 122 – 133.

[8] Garcıa V. M. A., Ruiz M. P. P., and Perez J. R. P(2010), “Voice interactive classroom, a service- oriented software architecture for speech-enabled learning”, Journal of Network and Computer Applications, 33 (2010) 603–610, Elsevier. pp. 603-610.

120

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

[9] Asha, S. and Chellappan, C.(2011), “Voice [19] Chen, H. J. (2010). Linking employees’ e- Activated E-Learning System for the Visually learning system use to their overall job outcomes: Impaired”, International Journal of Computer An empirical study based on the IS success Applications (0975 – 8887) Volume 14– No.7, model. Computers & Education, 55(4), 1628- February 2011 1639.

[10] Hsiu-Ju Chen and Chia-Hung Kao (2012), [20] Azeta, A. A., Ayo, C. K., Atayero, A. A. and “Empirical validation of the importance of Ikhu-Omoregbe, N. A. (2009b), A Case-Based employees’ learning motivation for workplace e- Reasoning Approach for Speech-Enabled e- learning in Taiwanese organizations”, Learning System”, 2nd IEEE International Australasian Journal of Educational Technology Conference on Adaptive Science & 2012, 28(4), 580-598 Technology (ICAST). December 14 – 16, 2009, Accra Ghana. ISBN: 978-1-4244-3523- [11] Blake, M. B. & Butcher-Green, J. D. (2009). 4, ISSN: 0855-8906. Agent-customized training for human learning performance enhancement. Computers & [21] Raghuraman M. B. (2004),”Design and Education, 53(3), 966-976. Implementation of V-HELP system – a voice- enabled web application for the visually [12] Asian productivity organization (2002), impaired. A Master’s project presented to the “Multimedia and e-Learning: A New Direction for faculty of the graduate college in the Productivity Promotion and Enhancement University of Nebraska in partial fulfillment of 2002. Report of the APO Seminar on requirements for the degree of Master of Multimedia for Productivity Promotion and Science. Major: Computer Science. August Enhancement (With Special Focus on e- 2004. Learning) Republic of China, 25–29 March 2002 [22] Voxeoprophecy(2003), “The Voxeo Prophecy Platform Free Download”, Available online at: [13] Prokopenko,J (1987) Productivity Management: www.voxeo.com/prophecy A Practical Handbook,, Geneva, International Labour Office. [23] Siemens G. (2003) “Open source content in education: Part 2 - Developing, sharing, expanding [14] Yesufu T. M. (2000). The Human Factor in resources”. national Development: Nigeria, Spectrum Books Limited, Ibadan, Nigeria. [24] Farrans, J. (2004), “COMMUNICATIONS OF THE ACM July 2004/Vol. 47, No. 7”, Piscataway, NJ [15] Akinyele, S. T. (2007). A Critical Assessment of Environmental Impact on Workers [25] Datamonitor (2006), “An Introductory Guide To Productivity in Nigeria. Res. J. Bus. Manage. Speech Recognition Solutions”, Understanding 1(1): 50-61. the technology, the vendors and the market Industry white paper published Datamonitor. [16] Osoba A. M. (1999). Productivity in Nigeria. (pp 1- 20) Proceedings of a National Conference, Productivity, Prices and Income Board, Ibadan. [26] ISO 9241-11, (1998). Ergonomic requirements for office work with visual display terminals [17] Umeh E. O. C., Usman GA (2000). Increasing (VDTs) - Part 11: Guidance on usability. Productivity in Nigeria: Proceedings from the 1st National Conference in Nigeria, National [27] Nielsen J., Usability 101 (2003): Introduction to Manpower Board, Lagos Usability, Alertbox, August, 2003. Online at www.useit.com/alertbox/20030825.html) [18] Wang, Y. S., Wang, H. Y. & Shee, D. Y. (2007). Measuring e-learning systems success in an [28] Sauro J. and Kindlund E.(2005), “A Method to organizational context: Scale development and Standardize Usability Metrics into a Single validation. Computers in Human Behavior, Score”, ACM, CHI, April 2-7, Portland, Oregon, 23(4),1792-1808. Retrieved online from USA, 2005. http://dx.doi.org/10.1016/j.chb.2005.10.006

121

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Authors’ Brief

Azeta, A. Ambrose. (Ph.D) is a lecturer in the Department of Computer and Information Sciences, Covenant University, Ota, Nigeria. He holds B.Sc., M.Sc. and Ph.D in Computer Science from University of Benin, University of Lagos and Covenant University respectively. His current research interests are in the following areas: Software Engineering, Algorithm Design and Mobile Computing. He currently lectures at Covenant University. He is a member of the Nigerian Computer Society (NCS) and Computer Professional Registration Council of Nigeria (CPN). .

Azeta I. Victor is a staff of National Productivity Center, Kaduna state office, Kaduna. He holds a B.Sc Political Science and M.Sc public administration from Edo State University and University of Calabar respectively. The current research interests include: Productivity, management, Administration and Finance, e-Government and Leadership style.

122

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Towards Sustainable Energy Development Using Virtual Power Plants

C.G. Monyei Department of Electrical and Electronic Engineering University of Ibadan Ibadan, Nigeria

ABSTRACT The Nigeria National Grid as it stands today cannot meet existing energy demand and with surging population explosion and an ever increasing energy demand, alternative energy sources and an efficient management of available energy is becoming the norm. While alternative energy sources like Renewable Energy Sources (RES) are stochastic in nature, they could with Advanced Metering Infrastructure (AMI) and an Energy Management System (EMS) be incorporated into the existing grid infrastructure to form a Virtual Power Plant (VPP). Complementing this VPP could be an increase in the reach of the grid and an interconnection of existing and would be incorporated components to form an intelligent system. It is envisaged that with such a structure in place, massive load shedding, frequent grid collapse and power outages could be reduced to minimum. This paper examines such system on a micro scale using Independence Hall and its surroundings all in the University of Ibadan as a case study.

Keywords- virtual power plant, renewable energy sources, advanced metering infrastructure

1. INTRODUCTION

The ultimate goal of any credible and legitimate government Corroborating Eddie, Helen Clarke [3] opines that energy is to ensure sustained improvement in the standard of living plays a central role in addressing two of the world’s greatest of the citizenry. To this end, developmental plans that will challenges: fighting poverty and addressing climate change. facilitate effective mobilization, optimal allocation and Smart energy policies have the potential to fight poverty and efficient management of scarce resources are evolved [2]. address climate change simultaneously.Nigeria has in recent The body charged with such varied responsibilities as power years been plagued with epileptic power supply consequent generation, transmission and distribution including metering upon corruption, old installations at major generating and collection of electricity levies etc. is the Power Holding stations, weak and obsolete transmission network grid and a Company of Nigeria (PHCN), now defunct as it were, poor distribution network. These have to a large scale disintegrated into three separate entities vis-à-vis the contributed to the poor power supply experienced within the Generation Company of Nigeria, Transmission Company of University of Ibadan Community that led to a near total Nigeria and the Distribution Company of Nigeria. overhauling of the energy supply and distribution network of the university. While the generation and distribution companies have been privatized, the transmission company still remains the In view of the above challenges that have plagued power property of the Nigerian Government. In the discharge of its generation and its delivery to its users, the University of responsibilities, PHCN is adjudged to be the largest single Ibadan has come up with some ideas of its own. consumer of natural gas in Nigeria with it accounting for The University of Ibadan as Nigeria’s foremost university has about 70 percent of the main fuel used in operating been doing a lot in recent times as regards Research and generating plants in Afam, Ughelli, Sapele and Egbin [4]. Development (R&D), and direct investments in modern With increased public outcry against gas flaring and the technologies to boost energy supply to and within the insidious environmental and energy consequences it university and to ensure a steady supply of power within its continues to generate against sustainable (energy) community. This has led to the purchase of generators, development in Nigeria [5], it has become imperative that installation of inverter systems within the halls of residence, more and better efficient ways of power generation and an replacement of underground distribution cables and effective use of (limited) available energy be evolved. As installation of prepaid meters in staff residential quarters, posited by Eddie Oldfield [7], energy poverty can be business premises etc.All these are being done by the addressed through the use of smarter technologies – which university in order for it to account for a greater part of its inform decisions we make as individuals, corporate citizens, energy consumption, distribution, create some level of energy policy makers and elected officials, with its exploration saving consciousness among users and overall reduce its honing on its ability to affect our choice of energy and wastage. Also, management has stepped up the installation of resulting economic impacts. Citing Gadonneix [1], Oldfield solar panels at strategic locations within the school further opines that while electricity is essential for our community to ensure steady illumination of walkways and heating, cooking and sundry other purposes, it is also linked streets across its length. to the world’s foremost social and environmental concerns.

123

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

2. THE PROBLEM As opined by A. Schäfer et al [9], a VPP is a collection of Due to the stochastic nature of the RES (photovoltaic power flexible consumers, which are grouped together and plants and wind plants), incorporating them into the controlled centrally. It is essentially an aggregation of university’s distribution grid firmware seems like an uphill Distributed Energy Resources (DERs) with an intelligent task hence they exist as independent distributed generation control allowing for their participation in an existing grid and storage units, serving just the vicinity within which they firmware. are installed. The disadvantage of this is that such DG units cannot have their stored energy accessed during peak periods, While it is not yet “uhuru” for the university’s management, neither is there any established communication protocol a critical appraisal of the ongoing trend is not encouraging between such DG units and the control room. (Uhuru is a a local slang with similar meaning to eureka. As a foremost institution of higher learning in the country, the Virtual Power Plant to the rescue installation of prepaid meters undermines the institutions As posited by A. Schäfer and A. Moser [9], one option of resolve at ensuring a greener and more efficient means of participation for dispersed generation is the aggregation in energy management and falls short of expected modern terms of a “virtual power plant”, thus representing one single evolving trend – smart meters. generation or demand unit. Optimizing such renewed drive that led to a near total overhauling of the university’s distribution network should Mette Petersen et al [8] posit that electricity is a so-called involve the installation of intelligent switches and the linking just-in-time product which means that it is instantly of every available DER including the installation of smart consumed at production. This implies that there has to be a meters at halls of residence, staff residential quarters and balance between electricity production and its consumption. business districts etc Integration of a VPP evens out the discrepancies between supply and demand via (short term) storage of energy or by voluntarily displacing consumption in time (Demand Side Management) [8].

A Block Porter’s Block A DG1 Lodge -200Ah, 24V (3 units) +56W, 440mA +910W, 7150mA

JCR, Buttery, walkway DG4 Block B -200Ah, 24V (3 units) Reserve/Backup -200Ah, 24V (3 units) +224W, 1760mA +910W, 7150mA

Block C Cafeteria DG3 +56W, 440mA -200Ah, 24V (3 units) +910W, 7150mA LEGEND - Estimated DER Block D Capacity. DG2 + Estimated block -200Ah, 24V (3 units) wattage and current. +420W, 3300mA DG – Distributed Fig 2: Independence Hall Generation. DER Layout.

3. ECONOMIC AND TECHNICAL EVALUATION

As surmised by Abimbola Odubiyi [6], demand will continue overemphasis on costly large centralised electricity supply to outstrip supply (in the short to medium term) which facilities is depriving some customers of the benefits that DG corroborates the earlier held assumption that with surging can provide. The case is no different for the University of population increase, electricity demand will continue to Ibadan which will continue to see an increase in its electricity increase. Abimbola Odubiyi [6] further opines that demand. 124

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Block A – DG1 - 200Ah, 24V (3 units) + 910W, 7150mA

Block D – DG2 - 200Ah, 24V (3 units) + 910W, 7150mA

Block C – DG3 - 200Ah, 24V (3 units) + 910W, 7150mA

JCR – DG4 - 200Ah, 24V (3 units) + 910W, 7150mA

Synchronizer

Control Room

LEGEND Please note that Block B has been left out and assumed to serve as a reserve/backup during an emergency – a scenario not considered in Controller/Switch this write – up.

Smartly Connected Loads

Block A Power Flow Path – DG1

Block C Power Flow Path – DG2

Block D Power Flow Path – DG3

JCR Power Path Flow – DG4

Synchronized/Augmented power Flow Fig 3: The proposed Independence Hall VPP Network Layout

In meeting this demand, it becomes imperative that cheaper, efficient, feasible, innovative and “greener” means of electricity generation be evolved. With abundant sunlight, good wind speeds, the presence of water body and isolated generating plants scattered within the institution’s vicinity including such pilot schemes as biomass etc., it becomes economical optimizing such varied options through a VPP. Meeting electricity demands (peak and off peak) through the installation of diesel driven generators is not only expensive (maintenance and operation cost-wise), but also environmentally unfriendly. In exploiting the proposed VPP project, we take advantage of its cost effectiveness, efficient energy management system and technical exposition – as it provides an avenue to exposing staff and students to the technical know-how of such a project.

125

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

Table 1: Average battery Life without VPP Table 2: Average Battery Life with VPP Block Mean battery life (hrs.) Block Min. battery life (hrs.) Max. Battery life (hrs.) A 11 A 19 21 B 11 B 19 21 C 13 C 19 21 D 16 D 19 21 JCR 21 JCR 19 21

15 Without VPP With VPP 10

0 Block A Block B Block C Block D JCR

Fig 3: A simple plot of Tables 1 and 2

25 without VPP with VPP 20

0 Block A Block B Block C Block D JCR

Fig 4: A bar chart distribution of Tables 1 and 2 126

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

5. CONCLUSION REFERENCES

As can be gleaned from fig 1, the shown layout displays no [1] Gadonneix, P. “Accessibility, Availability, interaction between the individual DERs. The loads (though Acceptability, Accountability – What do these not shown) are not smartly connected and energy four things have in common? Energy” in The Globe management systems are not in place. and mail, P.WECI, a special information feature (2010). Correcting the errors pertinent to fig1, the proposed VPP Network Layout (fig 2) displays an enhanced network replete [2] http://Boelle.de/download/internationalpolitik/Gipi with an EMS, real-time communicability, smartly connected 2.pdf loads, etc. Tables 1 and 2 display a summary of a survey carried out in [3] http://Web.undp.org/asia/pdf/nergyPlus.pdf the hall of the mean battery life time of existing DERs without a VPP and an extrapolation from a simple experiment [4] http://Worldenergy.org/documents/congresspapers/ using a VPP, EMS and regulated smartly connected loads. 70.pdf

Corroborating tables 1 and 2 are figs 3 and 4 which display a [5] Nwanya, S. C.” Climate change and energy visual representation of the data contained in tables 1 and 2. implicationsOf gas flaring for Nigeria”, retrieved from http://ijlct.oxfordjournals.org at While it is quick to point out that the case study might not University of Ottawa on August 19, 2012. provide the best avenue for such a pilot scheme to be tested (because of the nature of the DERs and their purpose), it is [6] Odubiyi, A. “Distributed Generation in Nigeria’s expedient to note that in effect a VPP aims at sourcing for New Energy Industry (2003)”, retrieved from power to meet peak demand through intelligent energy http://ieee.org on 11/11/2012. decisions and actions. While this aim has been satisfied by the experiment carried out (sourcing for power and intelligent [7] Oldfield, E. “Addressing energy poverty through decisions and actions), the performance of the system under smarter technology”, in Bulletin of Science, peak demand conditions has not been investigated. Future Technology & Society (2011), retrieved from work on this project hopes to investigate peak demand http://bst.sagepub.com/content/31/2/113/ on August performance and incorporate self- diagnostics and self- 14, 2012. healing capabilities. [8] Petersen, M., Bendtsen, J., Stoustrup, J. ” Optimal The global train as regards smart grids and better efficiency Dispatch Strategy for the Agile Virtual Power at delivering power to customers is en route and her Plant” in the 2012 American Control Conference, destination is within sight. The onus therefore lies on us in Fairmont Queen Elizabeth, Montreal, Canada. taking advantage of available cutting-edge technologies in complementing and supporting our already weakening [9] Schäfer, A., Moser, A. “Dispatch Optimization and national grid towards “Sustainable Energy Development in Economic Evaluation Of Distributed Generation in Nigeria”. a Virtual Power Plant”, retrieved from http://ieee.org on 31/10/2012.

END NOTES This paper was presented at the Annual Engineering Conference of the Nigerian Society of Engineers 2012

127

Vol 5. No. 6. Dec 2012 ISSN 2006-1781 African Journal of Computing & ICT

The African Journal of Computing & ICTs.

Call for Papers

INTRODUCTION PAGINATION FEES AND INFORMATION FOR SUBSCRIBERS AND ADVERTISERS The African Journal of Computing and ICT was established by the Nigeria Computer section of the Institute of Electrical Authors of accepted papers will be charged a pagination fee and Electronics Engineers (IEEE) Inc, New York, USA. Afr. of $150 (N20000). The subscription fee for yearly volumes of J of Comp and ICT publishes both online and print editions the journal to institutions and organizations is US 200 in April, August, December. The Journal solicits original (N30000 Naira) excluding postage. Copies of individual research articles from interested authors for her volumes can be obtained at a cost of $50 (N8000) upcoming editions. The goal of the journal is to contribute Correspondence on paper publications, book review, to rapid IT developments in Nigeria, Africa and the world subscription and advertisement should be directed to the Dr. through the publication of topical and relevant papers from Longe Olumide, Vice Chairman, IEEE Computer Section theoretic/academic and practical / industrial points of view. Nigeria and Managing Editor, African Journal of ICT. Submitted papers should contain substantial ideas which have Request for information pertaining to the Computer Chapter the potential of aiding an invention or a technological of the IEEE Nigeria Section should be sent to the Chair of the breakthrough for general and specialized use . Chapter, Prof. ‘Dele Oluwade [email protected]. SCOPE OF ARTICLES General information on the aim and objective as well as Papers are welcome from pure/applied computing and ICT, membership of IEEE, IEEE Computer Society and the IEEE and allied areas including but not limited to software Nigeria Section can be found on the following websites engineering, computer communication/networks, human respectively: http://www.ieee.org ; http:// computer.org ; computer interaction, data mining, discrete/data structures, http://www.ieee.org/nigeria. design & analysis of algorithms, web security, systems architecture, ubiquitous computing, systems design, information systems, bioinformatics, information management, information science and e-systems (e-health, e- learning, e-government).

The journal considers original research papers, case reports and review papers. Authors must clearly justify the importance of their submission to the readership of the journal. Prospective authors need not be a member of the (Computer Chapter) IEEE Nigeria Section before his/ her paper is published. All papers are peer refereed. The AJCICT Publication guideline can be found at www.ajocict.net.

128