A Symmetric Key Encryption for Data Integrity Varification Using Artificial Neural Network

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616 A Symmetric Key Encryption For Data Integrity Varification Using Artificial Neural Network

Monika P, Anita Madona

Abstract: An automatic programmed encryption framework dependent on neural networks beginning from the vital symmetric key model, the neural networks to make the conversation of the legal party more effective given less training steps, and to select the suitable hyper-parameters to enhance the statistical randomness to withstand distinguishing attack. In the current neural system, information trustworthiness isn't ensured, and Unique and arbitrary encryption keys are additionally fundamental for data honesty: an open key can't be acquired from an random key generator since it enables mediators to promptly assault the system and access delicate information. In this manner, so as to verify both data and keys, a plan ought to be utilized to deliver particular, mixed and arbitrary middle of the road encryption and decryption keys. Using the Automatic Artificial Neural Network (ANN) to introduce ideas such as authenticated encryption. ANN are the standards of finding the choice consequently by computing the proper parameters (loads) to cause the similarity of the framework and this to can be imperative to have the keys that used in stream figure cryptography to make the general framework goes to high security. The feature learned in Neural Network continues uncertain and we eliminate the chance of being a one-time pad encryption system through the test. With the exception of a few models wherever the assailants are too powerful, most models will be taught to stabilize at the training point. The newly suggested Elliptic Curve Diffie-Hellman (ECDH) based ANN important exchanges Automatic encryption systems are very powerful and flexible in resisting countless attacks. Further optimize the ANN to create the communication of the legal party much more effective given less training measures and how to select the suitable hyper-parameters to increase the statistical randomness to withstand distinguishing attack. This experiment carried out in MATLAB, together with an ANN, it used ECDH keys. The ANN replicates nature's randomness, where by a natural selection mechanism and natural system behavior a population of people adapts to their environments. The ANN produces a populace with high wellness esteem and the intervening cipher text utilized in encryption is this populace. The ANN will at that point utilize this intervening cipher to encode the underlying message. The ANN utilizes the calculation of mistake back propagation, which as its loads and biases utilizes its own key. Implementing concepts such as authenticated encryption by using the advanced ECDH with ANN would be intriguing to achieve high effectiveness in information integrity, time consuming, performance and precision.

Index Terms: ANN, Elliptic Curve Diffie-Hellman, Neural Network, Encryption ,Decryption ————————————————————

1 INTRODUCTION was actually sent by the sender.

1.1 Background 1.2 Research Aim and Objective In cryptography is the hiding data exercise and research. For Security solutions are commonly implemented through various secure communication, it is an important element. In addition kinds of networks in our daily digital communication. In the to protecting data from theft or alternation, cryptography can wireless network, complicated security protocols such as SSL / also be used to authenticate users. Cryptography can also be TLS, HTTPS and WPA have already been strongly used to described as converting information into a scrambled code safeguard our data transmission. However, owing to the safety that can be deciphered and transmitted through a government faults being found, the safety patches are continually being or private network. Cryptography utilizes two major data introduced to these safety alternatives. This study is very encryption styles or types; symmetrical and asymmetrical. For essential for implementing the safe system in digital encryption, symmetric encryption, or algorithms, uses the communication and internet or any system dealing with same key as for decryption. Other names are secret-key, information transportation and rejecting the attacker in an shared-key, and private-key for this form of encryption. For uncomplicated and cheap hardware system. encryption and decryption, asymmetric cryptography utilizes distinct encryption keys. In this case, a public or private end 1.3 Problem Faced During Data Transmission user on a network has a pair of keys; one for encryption and While using Cryptography for data transmission, we are facing one for decryption. These keys are marked or referred to as a some issue they are: private and a public key. There are certain particular safety conditions in the context of any application-to-application 1.3.1 Cryptographic Attacks interaction, including: The methods used by an opponent to break secured system • Authentication: The identification proof process. are known as cryptographic attacks. It is possible to classify • Privacy / Privacy: Ensure that no one, except the expected these assaults into two main classifications. recipient, can read the message.  Passive attacks • Integrity: Assure the recipient that the message obtained  Active Attacks was not changed from the original in any manner. Non-repudiation: a mechanism for proving that this message 1.3.1.1 Passive attacks In Passive Attacks, an attacker attempts to listen to the ———————————————— network link to obtain some data and attempts to break the  Ms.P.Monika is currently pursuing M.Phil (Computer Science) scheme based on the shared packets between sender and Research Scholar in Auxilium College (Autonomous), Vellore, Tamil receiver. Finding known plain texts is one instance of a Nadu, and India.Ph-88256 43506. E-mail: passive attack where an opponent monitors unencrypted traffic [email protected] and searches for sensitive data such as usernames or  Mrs. Anita Madona, Assistant Professor, Department of computer science, Auxilium College (Autonomous), Vellore, Tamil Nadu, India, passwords shared between communicating parties. Passive E-mail: [email protected] attacks are intended specifically to harm the sender and receiver while the system stays intact, which is the only reason 2132 IJSTR©2019 www.ijstr.org INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616 that most passive attacks are undetectable. Encryption should helping to keep confidence between exchanging sides. A be used widely for communication reasons to mitigate any digital signature that is not identical guarantees non- chance of reactive assaults such as snooping and analyzing repudiation. So, whenever a confidential data is transmitted data. over a network, it is digitally signed by the initial sender with its own private key and opened by the receiver using its public 1.3.1.2 Active attacks key. In effective assaults, in order to infect a laptop using Virus, Trojan horses or parasites, an attacker attempts to edit the 1.3.2.5 Authentication information being transferred over a network or within a Authentication guarantees that entry to protected and private scheme. Active assaults can be readily recognized, but they assets is available only to the correct individual. Authenticity are hard to avoid because while under assault, lawful users offers the recipient with evidence that the initial customer has have no power over their own scheme. Masquerade, Replay, sent data or text from his own email. Many methods are used Modification, and Denial of Service are four types of effective to provide validity such as OTP, usernames, passwords, facial assaults. recognition, iris scanning, and fingerprint scanning. Selecting a method or methods for authentication relies solely on the 1.3.2 Security Services criticality of a scheme. The International Telecommunications Union Telecommunications Standardization (ITU-T) defines five 1.4 Approach Chosen to Solve Problem safety facilities in Figure 1. In modern cryptography, data protection is accomplished through encryption. In multiple network protocols such as SSL / TLS and so on, symmetric key cryptography is primarily accountable for the real user data protection. Such encryption algorithms have always been designed as one of the most significant study goals, where work on intensive cryptanalysis was conducted to assess the safety margin. The research community is therefore busy solving the safety legislation depending on the outcomes of cryptanalysis. Recently, the concept was suggested to build the instant safety protection system centered on the artificial neural network. Instead, the encoding matrix, which is an interactive antenna network, is built by computer in an adversarial setting during the training Fig 1 Security Services phase. Compared to our present engineering theory, this is a

completely distinct strategy and could possibly alter our 1.3.2.1 Access Control knowledge of how the (inverted important) authentication The mechanism for access control is accountable for offering functions and what the scheme's safety obligation is. In this rights to various kinds of lawful customers operating in an job, the safety of the underlined system continues unexploited organisation or on a specific scheme. This system, according on the basis of several statistical designs is investigated. To to a person's classification, can provide or change permissions reinforce instant encryption schemes through the introduction linked to reading, writing and deleting a specific document. of much greater adversaries and the introduction of Elliptic Access control is also sometimes referred to as specific Curve Diffie-Hellman (ECDH) main transfers Automatic resource user limitation and can be supplied using usernames encoding systems are very powerful and versatile in withstand and biometric devices. countless assaults. Our findings indicated that the Artificial Neural Network-based safety alternatives with ECDH 1.3.2.2 Data Confidentiality algorithm are achieving elevated effectiveness in data integrity, Mechanism for data confidentiality guarantees that data is effort consuming, performance and precision. safely stored in a specific location and can only be transmitted by approved customers and is not revealed to an intruder. It is accomplished using encryption algorithms such as the Data 2 NARRATIVE OF ECDH WITH ARTIFICIAL Encryption Standard and the Standard for Triple Data NEURAL NETWORKS Encryption and Advanced Encryption Standard. 2.1 Artificial neural networks 1.3.2.3 Data Integrity An Artificial Neuron Network (ANN), commonly recognized as It refers to the confidentiality of information that has been Neural Network, is a computing system centered on biological physically stored on a hard drive or on a web server and neural network design and features. It is like an artificial protects it against any authorized modifications or deletions. It human nervous system for computer science data gathering, also requires charge of the data source where it was first handling and communicating. In a neural network, there are produced and subsequently transferred. Data Integrity system basically 3 distinct layers:- also guarantees that in the event of any disastrous 1. Input layer (through this coating all outputs are supplied circumstances such as storms and earthquakes, the data into the template) collected is not damaged. 2. Hidden Layers (More than one concealed stacks may be used to process the information obtained from the initial 1.3.2.4 Non repudiation strata) It is a system that ensures that the information originator is 3. Output layer (information produced accessible at the input unable to ignore the reality that he has sent the data, thus level after handling) 2133 IJSTR©2019 www.ijstr.org INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

2.1.1 Input Layer of concealed nodes as entry nodes have stronger predictive The input layer communicates with the neural network's outcomes. To chose to put that amount of concealed nodes to internal setting, which introduces a model. His task is only to the ultimate network for that purpose. handle all the variables. This entry is transmitted to the concealed levels described below. The entry section should be 2.2 General Theory of Artificial Neural Networks (ANNs) the situation for which the neural network is being trained. Artificial Neural Networks is just an synthetic neuronal Each entry neuron should be an independent variable that interconnection. ANNs know from past studies about the affects the performance of the neural network relationships between chosen inputs and outputs. ANNs also concurrently conduct their duties (i.e. simultaneous 2.1.2 Hidden Layer computing), making ANNs very quick. A typical ANN can The hidden layer is the neuron set that has activation function recognize and understand the relationship between the inputs attached to it and is an interface space discovered between and ##s of a multi-dimensional non-linear system (see Figure the surface of entry and the layer of production. His task is to 3). handle the outputs that his prior coating has acquired. So it is the coating that extracts the necessary characteristics from the entry information.

Fig 3: Non-linear multi-dimensional system

2.2.1 ANNs Architecture Mainly, ANNs are categorized into two categories based on the connection framework, feed-forward (i.e. from entry to output only) and recurrent (feed-back) networks.

2.3. Advantages of artificial neural network Neural networks can be used to obtain patterns and identify Fig 2. Shows a typical feed-forward network composed of one trends that are too complex to be observed by either concealed section of three nodes, five initial cells and one individuals or other software methods, with their notable generator. capacity to draw significance from complicated or imprecise information. One can think of a qualified neural network as an 2.1.3 Output Layer "specialist" scheme. Due to fresh cases of concern, this The neural network's output layer gathers and transmits the specialist can then be used to provide predictions and reply data in the manner it was intended to provide. The yield layer "what if" inquiries. Other benefits of artificial neural networks model can be drawn straight home to the entry layer. The (ANN) include: amount of input panel cells should be immediately linked to the sort of job performed by the cellular network. First imagine the (1) Adaptive teaching: an capacity to know how to perform planned use of the cellular network to determine the amount of duties depending on practice or original knowledge cells in the input unit. information.

2.1.4 Number of Neurons in each Layer (2) Self-organization: An ANN may establish its own There are three distinct levels in the suggested ANN. There is organisation or depiction of the data it gets during the training no question in the event of the entry sheet that the amount of period. nodes must be equivalent to the amount of features of the street lighting system, which is comprised of 13 distinct (3) Real Time Operation: ANN computations can be parameters in our situation. The amount of points must be performed in simultaneous and unique hardware systems are equivalent to the amount of required parameters in the event intended and produced using this capacity. of the yield, which in our event is one, the general uniformity. The amount of concealed layer layers, however, needs an (4) Redundant information coding fault tolerance: partial earlier research that can assist determine the finest setup. To destruction of the network leads to the corresponding determine the number of hidden nodes, experiments where degradation of performance. multiple configurations are compared or trial-and-error are the most common practice. However, some thumb guidelines (5) Another significant characteristic of artificial neural have been suggested. If we give regard to one concealed level networks is that, while each processor is very easy in terms of network research, we can discover distinct current rules such computing energy and storage, non-linear systems are as "2n+ 1," "2n," "n" or, "n/2," where n is the amount of entry adaptable. Artificial neural networks can therefore be used to points. While one of these proposals cannot be generalized estimate nonlinear models, an essential property to solve because none of these decisions operate for all issues, many issues in the real world. Artificial neural network designs several trials have shown how networks with the same amount ' adaptable parameters are the links that connect the 2134 IJSTR©2019 www.ijstr.org INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616 processors. This is comparable to "teaching" in biological An Artificial Neural Network is a network of many very easy neural networks due to modifications in the intensity of computers (modules), each of which may have a local storage neuronal links. (tiny number). The units are connected by numerical data carrying unidirectional communication channels. The devices 2.3.1 BENEFITS OF NEURAL KEY GENERATION. only work on their local information and obtain feedback via Benefits are the neutral generation of keys that can be the links. The development motive is what separates neural approved by the following:-1-In which a large number of networks from other mathematical methods: a neural network attackers have trained, and each new time step is multiplied by is a computing tool, either an algorithm or real hardware, each attacker to cover the 2k-1 possible internal whose development has been inspired by the development representations of the current output. 2-As dynamics precedes and working of natural bodies and elements. There are many staying effective perpetrators while removing the ineffective. distinct kinds of Neural Networks, each with distinct The Probabilistic Attack, in which the intruder attempts to weaknesses specific to their apps. Various networks ' skills pursue the likelihood of each weight component by calculating can be linked to their composition, dynamics, and techniques the local domain allocation of each entry and using the openly of teaching. recognized yield. 3-As stated in section two on the teaching concept that weight adjustment has been accomplished 3 ELLIPTIC CURVE DIFFIE-HELLMAN WITH ANN uniformly, so this randomization is unclear to the victim, these TECHNIQUES TO IMPROVE DATA INTEGRITY concepts contribute to a far-off adjustment region. Neural The ECDH important and Artificial Neural Network (ANN) synchronization can be used to build a cryptographic key- combination to decrease encryption and decryption period and exchange system because of this impact. Partners here profit also improve accuracy of data integrity. from shared communication, so a reactive intruder is generally unable to know in moment the code produced. 3.1 Elliptic curve cryptography: Elliptic curve cryptography (ECC) depends on the elliptic curve 2.4 THE DESCRIPTION OF SYSTEM hypothesis. Koblitz and Miller proposed the ECC idea, which is The scheme phases have several logic countries as shown in described briefly below, in order to plan public key Figure (4):-1) Initialize random weight measurements. cryptographic frameworks. Over a prime finite field Fp the 2) Run the previous measures until complete synchronization general type of elliptic curve E is: has been reached 3) Random X entry matrix generation. y2 = x3+ax+b (1) 4) Calculate concealed neuron characteristics. Where, a, b, fp and the D= 4a3 + 27 b20. Together with an 5) Calculate the output neuron price. extra point O, the locations on elliptic curve E over a given 6) Compare the scores of the two devices for tree parity. finite field Fp are called the location at infinity or the point at 7) There are other outputs: go to stage 2.1. null, which is referred to as: 8) Output are Same. A = {(x, y): x, y Fp, E (x, y) = 0} (1) ----- (2)

Let n be A's command so n g mod q = 0, where g is A's 2.4.1 Synchronization of neural networks generator. In relation, let A be a cyclic additive group under the Neural networks benefit from instances when both teacher and "+" stage, identified as P+0 = P, where P An11. The sum of pupil networks have N weights, the teaching method requires scalar points over A can be described as: the addition of N instances in order to gain generalization kP = P+P+……..+P (k times) ----- (3) skills. This implies that the pupil has attained some overlap If P, Q, and A is the base R of P+Q. Curve intercepts the row with the teacher after the teaching stage, their weight vectors going through P and Q at stage -R. -R reflects R in relation to have linked with each other. As a result, students can classify the x-axis. This situation is called the complement of the stage an input pattern that is not part of the training set. With the (Fig. 4). amount of training examples, the median evaluation mistake If two lines merge, i.e. P= Q, then R= P+P, becoming a tangent reduces. There are two distinct methods for coaching: batch at P, intersecting the -2P curve. The 2P picture on the y- and on-line instruction. All examples are stored in the first case coordinate's modified form is the consequence of P+P lying on and used to minimize the complete training error. Only one the E / FP line. This situation, as shown in Fig 5, is regarded fresh instance used per time step in the second situation and as level matching. then demolished it. Online training can therefore be regarded as a vibrant method: instructor generates a fresh instance at each phase that students use to modify their weights by a small quantity. ANNS created as generalizations of animal intelligence or cellular genetics mathematical designs. Based on the hypotheses: 1. There are many easy components called neurons that provide information. 2. Signals are carried through link connections between neurons. 3. Each link has weight connected with it. Which multiplies the transferred message in a typical neural net? 4. To determine its output signal, each neuron generally Fig. 4: Point addition uses an activation function that is nonlinear to its net input (total of weighted input variables).

popular buildings includes acclimations to the neuronal synaptic connections. The same can be said for ANNs as well.

3.3 Proposed approach: Neural cryptography relies on the effect of two genetic structures through popular teaching that are suitable for synchronization. Neural schemes obtain a typical information layout and determine their performance in each advancement of this internet approach. The two sensory circuits use their accomplice's inputs to alter their weights. This method Fig. 5: Point doubling contributes to weight matrices that are completely aligned. Neural network synchronization is actually a complicated vibrant method. Network weights conduct random walks motivated by an appealing and repulsive contest of stochastic powers. By cooperating with each other, two cellular networks can improve the appealing impact of their movements. A third network, however, which is trained only by the other two, clearly has a disadvantage because it cannot miss certain repulsive measures. Two bidirectional cellular structures can therefore extend the collaboration to foster the effect of their movements on each other. Anyway, a third system trained by the other two obviously has a disadvantage because it can't avoid some repulsive measures. Bidirectional synchronization is much quicker along these rows than unidirectional teaching. In particular, the application of the ANN should demonstrate high efficiency based on three metrics: time cost, performance and error rate. Figure 7 demonstrates the ANN encryption method with the ECDH button, which is how the client sends and encrypts text files.

Fig. 6: ECDH

Key transactions of Elliptic Curve Diffie-Hellman (ECDH): for the most portion, big percentage of data encryption / decryption needs the use of symmetric key (also known as secret key) cryptosystems due to their quicker calculations compared to public key cryptosystems. The ECDH main trade technique can be added to an elliptic curve to generate a secret key between two customers for one meeting. Assume that for secret key cryptography customers A and B need to agree on a secret key. Client A generates private key dA and public key PA=dAG, where G is the generator of elliptic curve. Client A gives PA to customerB. Similarly, private key dB and public key PB=dBG are produced by client B. Then Client B gives PB to Client A. Client B procedures dB (PA)= dAdBG upon submission of the message from client A. Client A Figures dA (PB)= dAdBG upon completion of the message from client B. At this stage, dAdBG, which is a location on the specified elliptic curve that serves as a typical secret key, can be used by customers A and B as shown in Fig. 6.

3.2 Artificial neural network: The ANN is a paradigm of data preparing that works as an organic sensory system, comparable to the data processing human brain. The main section of this paradigm is the data Fig. 7: ECDH+ANN training scheme framework, which consists of a big amount of considerably interconnected sections (cells) that work as one 3.3.1 Benefits of our Proposed System to deal with particular problems. The ANNs teach by example,  This suggested design is guaranteed to a large comparable to animals. An ANN implements a particular standard as the coaching entry adjusts its weights application through a teaching method, such as layout with qualified information. acknowledgement or demand for information. Learning in 2136 IJSTR©2019 www.ijstr.org INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

 Since our system provides the initial result rapidly and effortlessly, the simple code is readily encrypted and matrix code is produced with less moment.  It improves the design of the ANN by using many amounts of concealed parts and switches and thus offers greater safety for the cryptosystem.  Training cellular networks with their concealed weights are used as buttons. The magnitude of the button is tiny.  Our suggested system is quicker than the asymmetric authentication scheme using regular important, inverted key.  Data shops network code. So even if the data is hacked by the intruder, he cannot decrypt.  Data holders can rely heavily on their information, which is maintained securely through the network.

4 EXPERIMENTAL RESULTS It used ECDH buttons together with an ANN in this MATLAB- based experiment. The ANN replicates nature's randomness, where by a natural selection mechanism and artificial scheme Fig 8 Text encryption time over the cloud conduct a population of people adapts to their environments. The ANN produces a population with a strong fitness price and the medium cipher text used in encryption is this population. The ANN will then use this auxiliary cipher to encrypt the initial signal. The ANN uses the algorithm of mistake base propagation, which in the shape of its loads and biases utilizes its own button. Nevertheless, the text document will operate on three time-based situations:

AES+ NN ECDH+ANN ECDH+ANN's method of data security in ECDH-based network architecture proved and asserted that its structure was sufficiently safe for network information. However, the assessment of safety effectiveness requires further inquiry. As illustrated in Fig 8, In contrast with the AES+NN and AES situations, the ECDH with ANN achieves the highest interval rate because the NN itself requires time when given to the AES random. Uploading 10 text files will realize this research. Considering the decryption procedure of every one of the three situations, as appeared in the time interim in Fig.9 it can see enhancements in both the encryption and decryption times. The NN and ANN output and mistake frequency are problems in combining the two algorithms to apply to ECDH. Fig 9Text Decryption time over the cloud Thus, it conducts authentication and decryption in the output phase, as shown in Fig. Respectively, 10 and 11.

enhancing software or using stronger algorithms for instruction. Artificial Neural Network can in this manner be utilized as another strategy for information encryption and decryption. The motivation behind this exploration was to ensure communication crosswise over unbound networks and to shield private information from unlawful entry across government networks. This research has introduced a symmetric cryptography that produces both government and private keys for encryption and decryption procedures using ECDH with ANN (mistake home diffusion neural network). Utilizing ECDH with ANN, confidentiality of information, data integrity to prevent manipulation, authentication of senders and users and avoidance of the event where either receiver or sender denies any of the emails. It provided an ECDH-based encoding scheme with ANN. The last was used by a

Fig. 11: Performance Decryptions for AES + NN and continuously shifting key to assemble a powerful encryption ECDH+ANN system. The primary issue, therefore, is the moment of encryption. The moment taken as the primary metric, therefore, considering the ANN's encryption-decryption efficiency and error ratio over the ECDH.

5.2 FUTURE ENHANCE MENT As far as time devoured, execution and blunder rate in encoding and decoding text documents, the proposed technique with the expansion of the ANN with Genetic Algorithm (GA) will be superior to the ANN with ECDH.

BIOLOGRAPHY [1] Tope Komal , et al.; " Encryption and Decryption using Artificial Neural Network" , IARJSET , Vol. 2, Issue 4, April 2015 pp. 81-83 [2] Yousif Elfatih Yousif, Dr.Amin Babiker A/Nabi Mustafa, Dr.Gasm Elseed Ibrahim Mohammed" Review on Comparative Study of Various Cryptography

Fig 12 Error encryption Algorithms",IJARCSSE , Volume 5, Issue 4, April- 2015, pp. 51-55 The encryption and decryption processes for the NN and [3] William Stallings, “Cryptography and Network ECDH+ANN are also applied over the AES for the error metric. Security: Principles and practices, Dorling Kindersley In the Figure. 12, the ANN interval is 20 times greater than the (india) pvt ltd., 4th edition(2009). NN interval, which shows how the ANN tries to resolve the NN [4] Yousif Elfatih Yousif, Dr.Amin Babiker A/Nabi Mustafa, mistake level by means of a hybrid mixture. This study's " Cryptography Techniques based on Neural primary input is to improve safety by evaluating and testing the Networks",IJARCSSE , Volume 7, Issue 4, April- elevated precision efficiency of ANNs on customer information 2017, pp. 308-311 over oval surfaces. It then simulates with ECDH the suggested [5] Ajay Pal Singh , Parvez Rahi " Performance model of the neural network. To simulate the use of ECDH and Enhancement in Public key Cryptosystems for ANN mixed keys. The result showed feasible time, Security using RSA Algorithm " , IJARCCE , Vol. 5, performance and error-based results. Issue 11, November 2016 , pp. 359-362 [6] Oludele Awodele, Olawale Jegede" Neural Networks 5. CONCLUSION AND FUTURE ENHANCEMENT and Its Application in Engineering ", InSITE, 2009 [7] Andrej Krenker, Janez Bešter and Andrej Kos "

Introduction to the Artificial Neural Networks", 5.1 CONCLUSION Methodological Advances and Biomedical Artificial Neural Networks is a straightforward strong method Applications capable of emulating computing devices that are extremely [8] Moon, C. and B. Black, 2015. Application monitoring complicated. A comparison research was conducted between for cloud-based architectures. United States Patent two distinct architectures of the neural System and cited their Application No. 20150358391. merits / demerits. ANNs can be utilized to execute ttp://www.freepatentsonline.com/y 2015/0358391.html computational and sequential circuits that are very [9] Tirthani, N. and R. Ganesan, 2014. Data security in complicated. In data communication technologies, data cloud architecture based on diffie hellman and security is a primary issue. Using two methods, the use of elliptical curve cryptography. International Association ANN in the field of cryptography was investigated. A linear for Cryptologic Research. information encoding technique relying on a computer is https://eprint.iacr.org/2014/049.pdf intended. Also, a digital signal cryptography neural network is [10] Genkin, D., L. Pachmanov, I. Pipman and E. Tromer, evaluated. Better outcomes can be accomplished by 2138 IJSTR©2019 www.ijstr.org INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

2016. ECDH Key-Extraction via Low-Bandwidth Electromagnetic Attacks on PCs. In: Topics in Cryptology, Sako, K. (Ed.). Springer, New York, pp: 219-235. [11] Azad, S. and A.S.K. Pathan, 2014. Practical Cryptography: Algorithms and Implementations Using C++. CRC Press, USA., ISBN: 9781482228892, Pages: 365. [12] Lang, J. and R. Haakegaard, 2015. The Elliptic Curve Diffie-Hellman (ECDH). November 2015. http://cs.ucsb.edu/~koc/ ecc/project/2015Abstracts/Lang+Haakegaard.pdf [13] Saied, A., R.E. Overill and T. Radzik, 2016. Detection of known and unknown DDoS attacks using Artificial neural networks. Neurocomputing, 172: 385-393. [14] Shankar, K. and P. Eswaran, 2016. An Efficient Image Encryption Technique Based on Optimized Key Generation in ECC Using Genetic Algorithm. In: Artificial Intelligence and Evolutionary Computations in Engineering Systems, Dash, S.S., M.A. Bhaskar, B.K. Panigrahi and S. Das (Eds.). Springer, India, ISBN: 978-81-322-2656-7, pp: 705-714. [15] Gajra, N., S.S. Khan and P. Rane, 2014. Private cloud security: Secured user authentication by using enhanced hybrid algorithm. Proceedings of the International Conference on Advances in Communication and Computing Technologies, August 10-11, 2014, Mumbai, pp: 1-6.