Social Networking

SOCIAL NETWORKING

Also in this issue: > Harnessing the Power of Architectural Design Principles > Trust Me. Trust Me Not.

AUGUST 2016 www.computer.org 31st IEEE • 2017• INTERNATIONAL May 29-June 2, 2017 Paral lel and Buena Vista Palace Hotel Distributed Orlando, Florida USA Processing www.ipdps.org SYMPOSIUM

Orlando is home to a rich offering of indoor and outdoor attractions. Located a mile from Walt Disney World® and 4 miles from Epcot, the Buena Vista Palace Hotel is a 5-minute walk from Downtown Disney with a complimentary shuttle to all Disney Theme Parks and Water Parks. The sprawling Lake Buena Vista resort offers a full menu of amenities and family friendly activities as well as ideal meeting space for IPDPS 2017.

IPDPS 2017 CALL FOR PAPERS GENERAL CHAIR Authors are invited to submit manuscripts that present original unpublished Michela Taufer research in all areas of parallel and distributed processing, including the development (University of Delaware, USA) of experimental or commercial systems. Work focusing on emerging technologies and interdisciplinary work covering multiple IPDPS areas are especially welcome. PROGRAM CHAIR During submission, authors can indicate up to three subject areas that can come Marc Snir from any track. Topics of interest include, but are not limited to: (University of Illinois at Urbana Champaign, USA) • Parallel and distributed algorithms, focusing on topics such as: numerical and PROGRAM VICE-CHAIRS combinatorial parallel algorithms for analysis, machine learning and simulation; • Algorithms: parallel algorithms for accelerators, neuromorphic architectures, and other Pierre Fraigniaud (IRIF, France) non-traditional systems; algorithms for cloud computing; power-aware parallel algorithms; streaming algorithms; domain-specific parallel and distributed • Applications: algorithms; performance modeling and analysis of parallel and distributed Robert D. Moser (UT Austin, USA) ) algorithms; run-time algorithms and protocols for resource management, • Architecture: communication and synchronization on parallel and distributed systems.. Hillery Hunter (IBM Research, USA) & Robert Senger (IBM Research, USA) • Applications of parallel and distributed computing, including computational and data-enabled science and engineering, big data applications, parallel crowd • Software: sourcing, large-scale social network analysis, management of big data, cloud and Pavan Balaji (Argonne National Lab, USA) grid computing, scientific, biological and medical applications, and mobile • Multidisciplinary: computing. Papers focusing on applications using novel commercial or research Torsten Hoefler (ETH Zurich, Switzerland) architectures, big data approaches, or discussing scalability toward the exascale level are encouraged. KEYNOTES & TECHNICAL SESSIONS • Parallel and distributed architectures, including architectures for instruction-level and thread-level parallelism; petascale and exascale systems WORKSHOPS & PHD FORUM designs; novel big data architectures; special purpose architectures, including graphics processors, signal processors, network processors, media accelerators, COMMERCIAL PARTICIPATION and other special purpose processors and accelerators; impact of technology on Details at www.ipdps.org architecture; network and interconnect architectures; parallel I/O and storage systems; architecture of the memory hierarchy; power-efficient and green computing architectures; dependable architectures; and performance modeling IMPORTANT DATES and evaluation. October 18, 2016 Submit Abstract • Parallel and distributed software, including parallel and multicore programming October 23, 2016 Submit Paper languages and compilers, runtime systems, operating systems, resource Nov 28 – Dec 5, 2016 Review Feedback & Author Response management including, middleware for supercomputers, grids, clouds, and data January 8, 2017 Author Notification centers, libraries, performance modeling and evaluation, parallel programming paradigms, and programming environments and tools. Papers focusing on novel After January 8, 2017 Deadlines for Paper Submissions to Most Workshops software systems for big data and exascale systems are encouraged.

SPONSORED BY: IN COOPERATION WITH: ACM SIGARCH & SIGHPC IEEE Computer Society Technical Committee on Computer Architecture Technical Committee on Parallel Processing IEEE Computer Society Technical Committee on Distributed Processing IEEE Computer Society http://computer.org • +1 714 821 8380

Staff

Editor Manager, Editorial Services Content Development Lee Garber Richard Park

Senior Manager, Editorial Services Contributing Staff Robin Baldwin Christine Anthony, Lori Cameron, Carrie Clark, Chris Nelson, Meghan O’Dell, Dennis Taylor, Bonnie Wylie Director, Products and Services Evan Butterfield Production & Design Carmen Flores-Garvey, Monette Velasco, Jennie Zhu-Mai, Senior Advertising Coordinator Mark Bartosik Debbie Sims

Circulation: ComputingEdge (ISSN 2469-7087) is published monthly by the IEEE Computer Society. IEEE Headquarters, Three Park Avenue, 17th Floor, New York, NY 10016-5997; IEEE Computer Society Publications Office, 10662 Los Vaqueros Circle, Los Alamitos, CA 90720; voice +1 714 821 8380; fax +1 714 821 4010; IEEE Computer Society Headquarters, 2001 L Street NW, Suite 700, Washington, DC 20036. Postmaster: Send address changes to ComputingEdge-IEEE Membership Processing Dept., 445 Hoes Lane, Piscataway, NJ 08855. Periodicals Postage Paid at New York, New York, and at additional mailing offices. Printed in USA. Editorial: Unless otherwise stated, bylined articles, as well as product and service descriptions, reflect the author’s or firm’s opinion. Inclusion in ComputingEdge does not necessarily constitute endorsement by the IEEE or the Computer Society. All submissions are subject to editing for style, clarity, and space. Reuse Rights and Reprint Permissions: Educational or personal use of this material is permitted without fee, provided such use: 1) is not made for profit; 2) includes this notice and a full citation to the original work on the first page of the copy; and 3) does not imply IEEE endorsement of any third-party products or services. Authors and their companies are permitted to post the accepted version of IEEE-copyrighted material on their own Web servers without permission, provided that the IEEE copyright notice and a full citation to the original work appear on the first scree n of the posted copy. An accepted manuscript is a version which has been revised by the author to incorporate review suggestions, but not the published version with copy-editing, proofreading, and formatting added by IEEE. For more information, please go to: http://www.ieee .org/publications_standards/publications/rights/paperversionpolicy.html. Permission to reprint/republish this material for commercial, advertising, or promotional purposes or for creating new collective works for resale or redistribution must be obtained from IEEE by writing to the IEEE Intellectual Property Rights Office, 445 Hoes Lane, Piscataway, NJ 08854-4141 or pubs-permissions@ieee .org. Copyright © 2016 IEEE. All rights reserved. Abstracting and Library Use: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy for private use of patrons, provided the per- copy fee indicated in the code at the bottom of the first page is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923. Unsubscribe: If you no longer wish to receive this ComputingEdge mailing, please email IEEE Computer Society Customer Service at [email protected] and type “unsubscribe ComputingEdge” in your subject line. IEEE prohibits discrimination, harassment, and bullying. For more information, visit www.ieee.org/web/aboutus/whatis/policies/p9-26.html.

IEEE Computer Society Magazine Editors in Chief

Computer IEEE Micro IEEE MultiMedia Sumi Helal, University of Florida Lieven Eeckhout, Ghent Y ong Rui, Microsoft Research University IEEE Software IEEE Annals of the History Diomidis Spinellis, Athens IEEE Computer Graphics of Computing University of Economics and and Applications Nathan Ensmenger, Indiana Business L. Miguel Encarnação, ACT, Inc. University Bloomington

IEEE Internet Computing IEEE Pervasive Computing IEEE Cloud Computing M. Brian Blake, University of Maria Ebling, IBM T.J. Watson Mazin Yousif, T-Systems Miami Research Center International

IT Professional Computing in Science San Murugesan, BRITE & Engineering Professional Services George K. Thiruvathukal, Loyola University Chicago IEEE Security & Privacy IEEE Intelligent Systems Ahmad-Reza Sadeghi, Technical Daniel Zeng, University of Arizona University of Darmstadt www.computer.org/computingedge 1 AUGUST 2016 • VOLUME 2, NUMBER 8

THEME HERE 10 16 38 Social Telemedia: Engineering Spammers The Relationship Privacy in Social Are Becoming between Social Applications “Smarter” on Information and Twitter Networked Media 4 Spotlight on Transactions: Using Computer Intelligence for Depression Diagnosis and Crowdsourcing Bjorn Schuller and Jian Pei

8 Editor’s Note: Social Networking: Connecting People in New Ways 10 Social Telemedia: The Relationship between Social Information and Networked Media omar Niamut, Mu Mu, Spyros Denazis, and Nicholas Race 16 Engineering Privacy in Social Applications Pradeep K. Murukannaiah, Nirav Ajmeri, and Munindar P. Singh 22 How Socially Aware Are Social Media Privacy Controls? Gaurav Misra and Jose M. Such 26 Social-Sensed Multimedia Computing Peng Cui, Wenwu Zhu, Tat-Seng Chua, and Ramesh Jain 32 Social Resistance michael P. Friedlander, Nathan Krislock, and Ting Kei Pong 38 Spammers Are Becoming “Smarter” on Twitter chao Chen, Jun Zhang, Yang Xiang, Wanlei Zhou, and Jonathan Oliver 44 Whither Social Media Governance? virgilio A.F. Almeida, Danilo Doneda, and Yasodara Cordova

47 Harnessing the Power of Architectural Design Principles eoin Woods 51 Trust Me. Trust Me Not. Bill Horne 70 Software-Defined Networks Meet Cloud Computing David S. Linthicum 44 Departments Whither 6 Magazine Roundup 54 Computing Careers: Social Media Finding the Social-Networking Job You Want Governance? 56 Career Opportunities

Subscribe to ComputingEdge for free at www.computer.org/computingedge. SECTIONSPOTLIGHT TITLE ON TRANSACTIONS

In the study, the SimSensei virtual Using Computer human system interviewed both US military veterans and nonmilitary individuals to identify—based on Intelligence for their speech characteristics—those at risk for posttraumatic stress dis- order (PTSD) or major depression. All Depression Diagnosis had been “coded for depression and PTSD” by standard psychiatric ques- tionnaires. The findings support the and Crowdsourcing assumption that vowel reduction was indeed related to depression or PTSD diagnosis. This installment highlighting the work published in IEEE Computer Society journals comes from utomated voice-based diagnosis is attracting growing IEEE Transactions on A ective Computing and IEEE A interest in many healthcare- related areas. Given sufficiently ro- Transactions on Knowledge and Data Engineering. bust analysis, the efficiency of such a system would be unrivaled—a phone call to a recording service could suffice to make relatively inexpen- CAN VIRTUAL INTERVIEWERS vowel reduction— commonly observed sive automated diagnosis accessible “HEAR” REAL HUMANS’ among individuals with psychological to the general public. Overall, the DEPRESSION? disorders such as depression or neu- prospects of automatic voice-based Björn Schuller, University of Passau rological disorders such as Parkin- diagnosis are exciting, although we and Imperial College London son’s disease. To make diagnosis less must stay alert to its potential ethi- reliant on clinicians’ or caretakers’ cal implications. subjective impressions, the authors umans are reasonably good use SimSensei, a virtual human inter- at sensing when others are viewer that “listens” to patients’ voices BJÖRN SCHULLER is a professor depressed. In particular, during unprompted conversations. of complex and intelligent systems our hearing is attuned to They argue that this system could at the University of Passau, and Hthe voice-related changes often exhi- lead to more natural patient behav- a reader in machine learning at bited during depression. But could ior during interviews, more objective Imperial College London. Contact computers also be taught to “hear” evaluations, and reduced costs. him at [email protected]. depression in everyday conversations Scherer and colleagues suggest that outside a lab, even to the point of being employing virtual agents o ers clini- able to make a clinical diagnosis? cians increased control over the choice In their article “Self-Reported of stimuli. Because virtual humans can BIRD’SEYE VIEW OF Symptoms of Depression and PTSD be programmed a priori, they avoid the CROWDSOURCING Are Associated with Reduced Vowel predispositions or “impurities” that Jian Pei, Simon Fraser University Space in Screening Interviews,” Ste- human partners might introduce. The fan Scherer and his colleagues inves- authors also propose that virtual hu- tigate such an automated approach to man interviewers can reduce patients’ lphaGo recently beat a hu- diagnosis (IEEE Trans. A ective Com- stress, fear, and sense of being judged. man master at a very so- puting, vol. , no. , , pp. –). Ultimately, this encourages patients to phisticated board game. They focus on the acoustic changes speak more freely, thus providing richer But there are still many in vowel sounds—particularly on data for analysis. waysA in which computers can’t match 4 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 8 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE SECTIONSPOTLIGHT TITLE ON TRANSACTIONS our intelligence. Human and computa- timely reviews on crowdsourcing. analysis, planning, schema matching, tional intelligence, however, can com- Anand Chittilappilly and his col- mining, and spatial crowdsourcing. plement each other, and by combining leagues’ article takes a general view They also discuss crowdsourced opti- them in a smart manner, we can achieve of crowdsourcing systems and tech- mization and systems and identify ma- In the study, the SimSensei virtual more than by applying each individu- niques. Starting with a crowdsourcing jor research challenges. Using Computer human system interviewed both US ally. For example, through the Captcha framework, it reviews popular crowd- military veterans and nonmilitary mechanism, a smart algorithm and sourcing platforms, including those individuals to identify—based on many people together transcribed more that are standalone, metaplatforms, hese articles will appeal both Intelligence for their speech characteristics—those than million words with nearly general purpose, and specialized. The to advanced researchers wish- at risk for posttraumatic stress dis- percent accuracy. The key to such feats authors then discuss in depth three ma- Ting to catch up on the latest order (PTSD) or major depression. All is crowdsourcing, a fast-growing re- jor aspects of crowdsourcing: incentive crowdsourcing developments and to Depression Diagnosis had been “coded for depression and search and development area. design, task decomposition and assign- general readers interested in learn- PTSD” by standard psychiatric ques- The September issue of IEEE ment, and quality control. They use a ing the essentials of this fast-growing tionnaires. The findings support the Transactions on Knowledge and Data real scenario to deliberate on a series of eld. For investigators familiar with and Crowdsourcing assumption that vowel reduction Engineering will feature two survey technical issues, and conclude with im- the results surveyed, the articles also was indeed related to depression or articles by leading experts in this ex- portant future research directions. provide a valuable opportunity to re- PTSD diagnosis. citing research frontier: “A Survey of Guoliang Li and his colleagues’ ar- examine critical principles and re- General-Purpose Crowdsourcing Tech- ticle is data management–oriented and search challenges. This installment highlighting the work published niques” (A.I. Chittilappilly et al.), and focuses on three major crowdsourcing in IEEE Computer Society journals comes from utomated voice-based diag- “Crowdsourced Data Management: A properties: quality control, cost control, nosis is attracting growing Survey’’ (G. Li et al.). Though written and latency control. After a brief intro- JIAN PEI is a Canada Research Chair IEEE Transactions on A ective Computing and IEEE A interest in many healthcare- simultaneously and independently, duction, the authors provide a detailed in Big Data Science and a professor related areas. Given sufficiently ro- the two articles complement each review of crowdsourced operators in- in the School of Computing Science Transactions on Knowledge and Data Engineering. bust analysis, the efficiency of such a other in many aspects. Both individ- cluding selection, collection, join and at Simon Fraser University. Contact system would be unrivaled—a phone ually and together, they present com- entity resolution, ranking and sorting, him at [email protected]. call to a recording service could prehensive, concrete, insightful, and aggregation, categorization, skyline suffice to make relatively inexpen- CAN VIRTUAL INTERVIEWERS vowel reduction— commonly observed sive automated diagnosis accessible “HEAR” REAL HUMANS’ among individuals with psychological to the general public. Overall, the DEPRESSION? disorders such as depression or neu- prospects of automatic voice-based IEEE TALE 2016 Björn Schuller, University of Passau rological disorders such as Parkin- diagnosis are exciting, although we and Imperial College London son’s disease. To make diagnosis less must stay alert to its potential ethi- 7-8 December 2016, Dusit Thani Bangkok Hotel, Bangkok, Thailand reliant on clinicians’ or caretakers’ cal implications. KeepingSpecial Track on Computing Education subjective impressions, the authors umans are reasonably good use SimSensei, a virtual human inter- “Innovations for Computing Education” at sensing when others are viewer that “listens” to patients’ voices BJÖRN SCHULLER is a professor Co-ChairsYOU at the SPONSORED BY depressed. In particular, during unprompted conversations. of complex and intelligent systems Prof. Sorel Reisman, California State University our hearing is attuned to They argue that this system could at the University of Passau, and Dr. Henry Chan, The Hong Kong Polytechnic University Hthe voice-related changes often exhi- lead to more natural patient behav- All the Knowledge a reader in machine learning at bited during depression. But could ior during interviews, more objective ComputingCenter is undergoing big changes. Nowadays, computers are not only desktop machines but alsoYou lightweight Need— notebooks Imperial College London. Contact inside your bags and smart devices inside your pockets. With the popularity of cloud technologies, computing can now be computers also be taught to “hear” evaluations, and reduced costs. On Your Time. him at [email protected]. delivered as utility services. Furthermore, computing knowledge is changing so fast that what students learn today will likely depression in everyday conversations Scherer and colleagues suggest that Learn something new. outside a lab, even to the point of being employing virtual agents o ers clini- becomeof obsolete Technology tomorrow. There are also many emerging technologies and methodologies for computing education. In view able to make a clinical diagnosis? cians increased control over the choice of such trends and developments, there is a need to study how computing education should changeTry Computer and innovate, fromSociety both In their article “Self-Reported of stimuli. Because virtual humans can BIRD’SEYE VIEW OF curricular and technology perspectives. Under the theme “Innovations for Computing Education”, eLearningthis special track today! is spon- Symptoms of Depression and PTSD be programmed a priori, they avoid the CROWDSOURCING soredComputer by the IEEE Computer SocietySociety. It provides eLearning an interactive forum for educators and researchers to exchange views, ideas and experiences on computing education. There will also be a panel on Open Education Resources (OER) for computing educa- Are Associated with Reduced Vowel predispositions or “impurities” that Jian Pei, Simon Fraser University Space in Screening Interviews,” Ste- human partners might introduce. The tion. You are invited to submit papers for this special track. fan Scherer and his colleagues inves- authors also propose that virtual hu- Important Dates: 20 Aug 2016 - Paper Submission Deadline tigate such an automated approach to man interviewers can reduce patients’ lphaGo recently beat a hu- Stay relevant with 20 the Sept IEEE 2016 Computer - Review Society Outcomes diagnosis (IEEE Trans. A ective Com- stress, fear, and sense of being judged. man master at a very so- puting, vol. , no. , , pp. –). Ultimately, this encourages patients to phisticated board game. More at www.computer.org/elearning 20 Oct 2016 - Final Paper Due They focus on the acoustic changes speak more freely, thus providing richer But there are still many Further Information: Please visit http://www.tale-conference.org/tale2016/cfp_SSCE.php in vowel sounds—particularly on data for analysis. waysA in which computers can’t match www.computer.org/computingedge 5 8 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE JULY 2016 9 CS FOCUS

Magazine Roundup

the challenges for signaling in emerging network architectures, as well as ongoing stan- dardization eff orts.

Computing in Science & Engineering

CiSE’s July/August 2016 special issue deals with the application he IEEE Computer IEEE Software of discrete modeling and sim- Society’s lineup of 13 ulation tools to problems from Tpeer-reviewed technical Articles in IEEE Software’s July/ diff erent fi elds, such as physics magazines covers cutting-edge August 2016 special issue on and engineering. The articles topics ranging from software software quality address mat- examine discrete models’ com- design and computer graphics ters such as quality improve- puting abilities and principles to Internet computing and secu- ment from the customer’s with a focus on their expressive rity, from scientifi c applications perspective; metrics for software dynamics, emergent computa- and machine intelligence to security, privacy, and depend- tion, and inherent parallelism, cloud migration and microchip ability; and the eff ectiveness of which make them suitable for manufacturing. Here are high- pair programming. high-performance computing. lights from recent issues. IEEE Internet Computing IEEE Security & Privacy Computer Various protocols provide sig- For decades, code obfusca- Future innovations in com- naling services in IP-based tion has been used to protect puting technologies—such as networks. “Signaling in IP- software from reverse engi- wearable devices, robots, and Based Networks: Retrospect neering. However, there’s been machine intelligence—require and Prospect,” from IEEE Inter- little proof that obfuscation wholesale transformation of net Computing’s May/June 2016 protects software from being user interfaces. This is the focus issue, looks at the development “deobfuscated.” “Cdoe Obofsu- of Computer’s July 2016 special of the Internet Engineering Task caitn: Securing Software from issue on user-interface design Force’s Next Steps in Signal- Within,” from IEEE S&P’s May/ for the 21st century. ing protocols. It also discusses June 2016 issue, discusses

6 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE how computer scientists and engi- IEEE MultiMedia The environment provides tools neers are working to change code that let people augment and con- obfuscation from a wobbly art to a With the advent of the digital age trol their home’s capabilities. sound methodology. and the growing popularity of portable audio players, interest has IT Professional IEEE Cloud Computing increased in software and hardware that helps music producers There are many new Internet of Cloud computing continues to grow and distributors enhance their cat- Things (IoT) applications, particu- in complexity. Autonomic comput- alogs. One of major music labels’ larly in healthcare. The technology ing helps deal with this complexity main concerns is protecting their can signifi cantly improve patients’ by enabling the self-management intellectual-property rights by pre- well-being while requiring rela- of systems and applications, as venting fi le sharing.Managing “ tively few resources. However, its addressed in IEEE Cloud Comput- Intellectual Property in a Music benefi ts have been overhyped. ing’s May/June 2016 special issue Fruition Environment,” from IEEE Furthermore, these technologies on autonomic clouds. MultiMedia’s April–June 2016 issue, could disassociate medical provid- proposes an innovative approach ers from patients. In “The Internet IEEE Computer Graphics and to this challenge. of Things in Healthcare: Poten- Applications tial Applications and Challenges,” IEEE Annals of the History of from IT Pro’s May/June 2016 issue, In addition to enabling students to Computing the authors review some of the obtain their PhDs, doctoral dis- IoT’s most promising healthcare sertations also help them enhance Some concepts and components applications and the signifi cant their reputations and fi nd desirable of Amiga home-computer tech- obstacles the approach faces. jobs. However, the process of pro- nology, initially released 30 years ducing successful dissertations— ago, have had staying power with IEEE Micro topic selection, research, writing, developers and users despite the revising, and defending—is com- original manufacturer’s demise. Datacenters are quickly becoming plex. This process is analyzed in According to “Retrospective the platform of choice for modern “The PhD Thesis Deconstructed,” Computing and Consumer-Led applications. To improve future from IEEE CG&A’s July/August Development,” from IEEE Annals’ processor performance, “Profi ling 2016 issue. April–June 2016 issue, this is an a Warehouse-Scale Computer,” example of continued technology from IEEE Micro’s May/June 2016 IEEE Intelligent Systems development by a community of issue, studied more than 20,000 invested users. Google machines to understand “The Role of Data Science in Web how datacenter software uses the Science,” which appears in IEEE IEEE Pervasive Computing hardware. Intelligent Systems’ May/June 2016 issue, notes that data science, as The authors of “A First-Person Computing Now an approach to use within Web Experience with End-User Devel- science research, off ers opportu- opment for Smart Homes,” from The Computing Now website nities to uncover trends in large IEEE Pervasive Computing’s April– (http://computingnow.computer Web-based datasets. The arti- June 2016 special issue on domes- .org) features up-to-the-minute cle outlines these two subject tic pervasive computing, present computing news and blogs, areas’ similarities and diff erences their experience with a prototype along with articles ranging from to demonstrate their important end-user development environ- peer-reviewed research to opinion relationship. ment deployed in their own homes. pieces by industry leaders. www.computer.org/computingedge 7 EDITOR’S NOTE

Social Networking: Connecting People in New Ways

s social networking becomes more pop- knowing who is friends with whom, inferring the ular, it is also becoming more varied in closeness of two people’s research areas just by A its approach. Social networking is being observing coauthor relationships, and predicting applied in new and complex ways, utilizing diff er- the importance of new roads just by looking at a ent technologies in the process. However, keeping current traffi c network. up with these changes can be a challenge. This “Spammers Are Becoming ‘Smarter’ on Twit- ComputingEdge issue looks at how social net- ter,” from IT Professional, analyzes how, although working is coping and what its future holds. researchers develop various approaches to detect Social telemedia is a cross between social net- Twitter spam, spammers thwart their eff orts with works and networked media that lets users cap- more complex strategies. ture and share live events collaboratively on mobile Computer’s “How Socially Aware Are Social Media devices. This is the topic of Computer’s “Social Privacy Controls?” notes that although social-media Telemedia: The Relationship between Social Infor- sites are key online-communication mediators, their mation and Networked Media.” privacy controls are often not fully socially aware. Privacy is crucial to social-application users. ComputingEdge articles on topics other than With this in mind, IEEE Internet Computing’s “Engi- software include the following: neering Privacy in Social Applications” introduces Danio, a methodology based on the modeling of • “Harnessing the Power of Architectural Design users’ expectations. In preliminary evaluations, it Principles,” from IEEE Software, discusses simplifi ed the development of social applications how these principles clearly defi ne the neces- and saved time during implementation and testing. sary constraints on a system’s design without The authors of “Social-Sensed Multimedia Com- prescribing all details. puting,” which appears in IEEE MultiMedia, propose • IEEE Cloud Computing’s “Software-Defi ned Net- a social-sensed multimedia computing paradigm works Meet Cloud Computing” reviews soft- and advocate for integrating social-network and ware-defi ned networking’s advantages; explains social-media data with multimedia computing tasks. its business and technology opportunities; and In Computing in Science & Engineering’s “Social looks at the potential for it to live in the cloud, pro- Resistance,” the authors investigate approaches viding an on-demand networking capability that to overcoming challenges such as measuring how could off er better and cheaper approaches to net- close two social-network participants are just by working.

18 October 2016 | Mountain View, CA

Why Attend Rock Stars of Pervasive, Predictive Analytics?

Want to know how to avoid the 4 biggest problems of predictive analytics – from the Principal Data Scientist at Microsoft? Take a little time to discover how predictive analytics are used in the real world, like at Orbitz Worldwide. You can stop letting traditional perspectives limit you with help from Mashable’s chief data scientist. Sameer Chopra Juan Miguel Lavista Haile Owusu Come to this dynamic one-day GVP & Principal Data Scientist Chief Data Scientist Chief Analytics Officer Microsoft Mashable symposium. Orbitz Worldwide

www.computer.org/ppa SOCIAL COMPUTING

Social Telemedia: The Relationship between Social

streams to a rich medium for shar- Information and ing audiovisual content. More and more users are capturing and sharing video content, which can largely Networked Media be attributed to the increased availability of video-recording capabilities on consumer devices such as Omar Niamut, TNO smartphones and their integration Mu Mu, University of Northampton with online content-sharing plat- Spyros Denazis, University of Patras forms. As a result, amateur video capturing has evolved from a per- Nicholas Race, Lancaster University sonal hobby to citizen journalism. New crowdsourced live mobile Social telemedia is a cross-breeding of streaming applications, such as Meerkat (http://meerkatapp.co) and social networks and networked media that Periscope (www.periscope.tv), have experienced explosive growth in allows users to capture and share live events popularity in the past few years. We collaboratively on mobile devices. refer to this community-centric, dig- itally based ecosystem as social telemedia, de ned as a cross-breeding of social networks and networked media. Social telemedia enhances ith the growing popularity of social media current social media practices, such as sharing photos sites, online video services, and smart- and videos, and it ourishes on networked middleware phones, content consumers are recording, frameworks and systems that e ciently combine so- editing, and broadcasting their own sto- cial informatics and content delivery. Social informat- Wries. Social media sites have evolved from text and photo ics loosely refers to any digitized information that’s 10 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 92 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE EDITOR CHRISTIAN TIMMERER Alpen-Adria-Universitat Klagenfurt; SOCIAL COMPUTING [email protected]

generated or exchanged in the context of social networking, and content delivery is the communication medium through which social informatics are exchanged. Examples of social informatics are social relationships in peer-to-peer (PP) systems, content Search and delivery powered by context and user recommendation intelligence, and context-aware net- Story editing and sharing working and communications. Media distribution and synchronization STEER: SOCIAL TELEMEDIA Social Telemedia: ENVIRONMENT FOR EXPERIMENTAL RESEARCH Capturing live video footage, sharing Figure 1. Social Telemedia Environment for Experimental Research (STEER) system. The Relationship highlights, and circulating narratives have become an important part of par- ticipant activities at live social events. User community External Keywords and Amateur video recordings with unique Social context social identi ed events networks between Social personal perspectives tend to elicit Location, more vibrant and immersive commu- related events Social graph nity experiences than professional and streams to a rich medium for shar- commercial recordings. User applications Story editing “Edit by reference” (mobile, Web, etc.) APIs story authoring Information and ing audiovisual content. More and Social telemedia systems leverage Statistics more users are capturing and shar- social media and advanced network- User interaction Ingest, annotation Transparent caching EDL conformation and system logs service ing video content, which can largely ing technologies to enable novel ap- be attributed to the increased avail- plications and new user experiences. Multimedia content Rich chunk Socially aware media Networked Media processing distribution ability of video-recording capabil- A social telemedia application iden- identi er ities on consumer devices such as ti es and assembles media content Omar Niamut, TNO smartphones and their integration produced by event participants or by Figure 2. Collaborative Storytelling application. EDL: edit-decision list. Mu Mu, University of Northampton with online content-sharing plat- automated systems from various rel- Spyros Denazis, University of Patras forms. As a result, amateur video evant sources. The content, ranging capturing has evolved from a per- from video clips recorded on personal Collaborative Storytelling indexing so that user- generated content Nicholas Race, Lancaster University sonal hobby to citizen journalism. cameras (with audio and superim- The Collaborative Storytelling appli- can be better shared and discovered. New crowdsourced live mobile posed comments) to segments of live cation enables communities to col- The story- authoring engine uses a Social telemedia is a cross-breeding of streaming applications, such as TV programs, is then distributed to laboratively author and edit video unique and lightweight “edit by refer- Meerkat (http://meerkatapp.co) and wider audiences with the help of smart narratives of a live social event., ence” design to enable collab orative on- social networks and networked media that Periscope (www.periscope.tv), have tools and networking infrastructures. Figure provides an overview of the line story editing. It uses edit- decision experienced explosive growth in We developed STEER (Social Tele- application and the STEER system’s lists (EDLs) to de ne time-addressable allows users to capture and share live events popularity in the past few years. We media Environment for Experimental underlying components. A Web portal references of user-generated content so collaboratively on mobile devices. refer to this community-centric, dig- Research), a system to facilitate social and a mobile application provide the that users’ stories can be made by ma- itally based ecosystem as social tele- telemedia. The system is depicted in user- facing elements including video nipulating hierarchical references to media, de ned as a cross-breeding Figure along with its applications, capture, sharing, searching, and story static video objects without any com- of social networks and networked Collaborative Storytelling and Live authoring. The user applications also plex video rendering. The social- context media. Social telemedia enhances Augmented Broadcast (LAB). The mo- capture live metadata such as location integration function monitors related ith the growing popularity of social media current social media practices, such as sharing photos bile application allows users to cap- information, which is used to provide events and trends on social media sites sites, online video services, and smart- and videos, and it ourishes on networked middleware ture video for asynchronous storytell- personalized media experiences. The to improve the user experience in tag- phones, content consumers are recording, frameworks and systems that e ciently combine so- ing and to stream live user-generated multimedia processing function en- ging user creations and to discover editing, and broadcasting their own sto- cial informatics and content delivery. Social informat- video to augment professional broad- compasses operations such as transcod- shared stories related to nearby trend- Wries. Social media sites have evolved from text and photo ics loosely refers to any digitized information that’s casting of the same event. ing, chunking, content ana lysis, and ing topics. www.computer.org/computingedge 11 92 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE MAY 2016 93 SOCIAL COMPUTING

in timeline alignment of mobile Time server communication Player-side streams, broadcast streams, and so- synchronization 3 desynchronized feeds STEER server communication Synchronized video streams cial content. P2P delivery is employed for efficient delivery of synchronized Player P2P video streams.8

NETWORKING TECHNOLOGIES STEER session The STEER system and the related so- server cial telemedia applications leverage the following social media and ad- Ethernet Ethernet Ethernet vanced networking technologies. Time server Adaptive Event Profiler With the Adaptive Event Profiler (AEP),9 one can follow news connected to an event on social media sites such as Facebook and Twitter. The profiler Ingest UGC Ingest Ingest 1 camera 0 camera 1 allows for emerging-event detection through event tracking on social me- Wi-Fi / LTE HD-SDI HD-SDI dia, and for personalization through ranking and clustering social media posts. Use cases include suggesting related terms to an event for tagging videos, following social media mes- Figure 3. Live Augmented Broadcast (LAB) application. P2P: Peer to peer; UGC: user- sages around an event in real time, generated content. and creating a social media dataset about a topic. The AEP also allows the retrieval of social media data in a significantly higher volume and degree of Repeat regularly during event to pick accuracy than simply following a sin- up new keywords gle keyword. The AEP, shown in Figure 4, runs Event title Initial search Set of Twitter Derive search from TV as a Web service on any cloud envi- keywords search results keywords to guide add ronment and uses Twitter as a source Rule-based Search / stream search Relevance of textual content to find related key- preprocessing keywords using Twitter API feedback words. It retrieves significantly more interesting and relevant Tweets than Write tweets to database following a single keyword, without introducing too much noise in the dataset. The API allows users to list all Figure 4. Adaptive Event Profiler (AEP), an innovative state-of-the-art social media events, create and stop events, check analysis platform. event details, and open specific Twit- ter streams. During public events, tweets are Live Augmented Broadcast home. An analysis of social network continually posted with information Figure 3 shows the LAB application, messages is performed to retrieve such as news about the participants and which combines professional broad- and show the most relevant posts in event outcomes. Facebook event pages casts of live events with live user-gen- sync with a video. By providing addi- can be tracked, and images posted on erated content from mobile devices.7 tional event-related social and video the event page can be retrieved and Event participants make live record- content, viewers can enjoy an aug- shown in the LAB application. ings with mobile devices, which can mented view of live events. Using a be streamed and watched in sync common clock synchronized via the Media Asset Referencing System with live broadcasts by viewers at Network Time Protocol (NTP) results One unique feature of the STEER

12 ComputingEdge August 2016 94 COMPUTER WWW.COMPUTER.ORG/COMPUTER SOCIAL COMPUTING

in timeline alignment of mobile sys tem is the “edit by reference” fea- Time server communication Player-side streams, broadcast streams, and so- ture. Instead of physically editing and synchronization 3 desynchronized feeds STEER server communication Synchronized video streams cial content. P2P delivery is employed rendering a video object, which con- for efficient delivery of synchronized sumes computational and network Player P2P video streams.8 re sources for every edit, a story can be composed by creating a list of references NETWORKING to existing video objects. This feature, TECHNOLOGIES called the Media Asset Referencing STEER session The STEER system and the related so- System (MARS), is a media manage- server cial telemedia applications leverage ment system that recognizes time-ad- the following social media and addressable media assets within chunked Ethernet Ethernet Ethernet vanced networking technologies. media objects and generates new media Time server objects on the fly by combining only the Adaptive Event Profiler segments that represent time periods With the Adaptive Event Profiler that fall within or overlap with those (AEP),9 one can follow news connected assets. MARS’s core is the management to an event on social media sites such of EDLs, script-based expressions of Figure 5. LAB application showing synchronized video and Twitter feed. as Facebook and Twitter. The profiler a media object’s composition. MARS Ingest UGC Ingest Ingest 1 camera 0 camera 1 allows for emerging-event detection creates a time- addressable descriptive through event tracking on social me- view of the internal structure of media communities. We demonstrated the tweets were typically generated by Wi-Fi / LTE HD-SDI HD-SDI dia, and for personalization through objects using frame- accurate presenta- feasibility of these applications and viewers all over the world who were al- ranking and clustering social media tion and timecode navigation, so that the underlying STEER system through ready combining their TV-viewing ex- posts. Use cases include suggesting virtual operations visually assisted by several experiments. perience with their social network use. related terms to an event for tagging keyframes such as cut and fast-forward The experiment considered a number videos, following social media mes- can be carried out without having to 2014 World Rowing of research questions: Figure 3. Live Augmented Broadcast (LAB) application. P2P: Peer to peer; UGC: user- sages around an event in real time, physically examine the corresponding Championships generated content. and creating a social media dataset media objects. During the 2014 World Rowing Cham- › How can we design an event pro- about a topic. The AEP also allows the pionships in Amsterdam, we assessed filer that generates a set of query retrieval of social media data in a sig- Hybrid media synchronization the LAB application with a focus terms to retrieve relevant and nificantly higher volume and degree of Media synchronization technologies on crowdsourced live mobile video interesting tweets during Repeat regularly during event to pick accuracy than simply following a sin- for live event coverage enable the streaming over 4G and synchronizing an event? up new keywords gle keyword. seamless combination of professional user-generated video with professional › How does the AEP perform The AEP, shown in Figure 4, runs broadcast and amateur videos shot broadcasts (see Figure 5). We recorded in generating relevant and Event title Initial search Set of Twitter Derive search from TV as a Web service on any cloud envi- by the crowd, as well as synchroniza- content from a number of different interesting tweets, compared keywords search results keywords to guide add ronment and uses Twitter as a source tion with social media feeds such as positions around the venue to obtain to a traditional tool based on a Rule-based Search / stream search Relevance of textual content to find related key- Twitter and Facebook streams. In the real-world results. We found that it’s simple event hashtag? preprocessing keywords using Twitter API feedback words. It retrieves significantly more LAB application, all media is time syn- quite difficult to create user-generated › What is the relation between interesting and relevant Tweets than chronized, re-creating an immersive recordings that appeal to a broad au- relevancy of and interest in Write tweets to database following a single keyword, without user experience. Content can be in- dience. For example, the crowd is only tweets, also depending on the introducing too much noise in the gested by a local node or in the cloud. of interest during key moments, and AEP settings? dataset. The API allows users to list all Ingest nodes work together to achieve other scenes are typically not related to Figure 4. Adaptive Event Profiler (AEP), an innovative state-of-the-art social media events, create and stop events, check frame-accurate media timestamping. the event. Added value mostly occurs During the experiment, two ses- analysis platform. event details, and open specific Twit- In other words, all media sources can in the cases of limited professional cov- sions with 10 participants took place ter streams. insert timestamps in the captured me- erage, clear social context, and crowd- at two different locations. All par- During public events, tweets are dia streams using common synchro- sourced mobile journalism. ticipants watched a live broadcast of Live Augmented Broadcast home. An analysis of social network continually posted with information nized clocks. Frame-accurate synchro- men’s 10,000 m (session A) and wom- Figure 3 shows the LAB application, messages is performed to retrieve such as news about the participants and nized playback is achieved by using 2014 Winter Olympics en’s 5,000 m (session B) speed skating which combines professional broad- and show the most relevant posts in event outcomes. Facebook event pages these timestamps. During the 2014 Winter Olympics in finals, with approximately 25 min- casts of live events with live user-gen- sync with a video. By providing addi- can be tracked, and images posted on Russia, we assessed the interest in utes of viewing time. Tweets shown erated content from mobile devices.7 tional event-related social and video the event page can be retrieved and EXPERIMENTS and relevancy of social media posts to the participants were evaluated on Event participants make live record- content, viewers can enjoy an aug- shown in the LAB application. Social telemedia applications provide on Twitter during a live TV broadcast. a seven-point Likert scale for interest ings with mobile devices, which can mented view of live events. Using a users with novel and exciting opportu- The AEP retrieved and filtered selected and relevancy. be streamed and watched in sync common clock synchronized via the Media Asset Referencing System nities to capture live events on mobile tweets during a series of live broad- When shown each tweet, partici- with live broadcasts by viewers at Network Time Protocol (NTP) results One unique feature of the STEER devices and share their stories with casts of speed skating events. These pants gave feedback on whether they

www.computer.org/computingedge 13 94 COMPUTER WWW.COMPUTER.ORG/COMPUTER MAY 2016 95 SOCIAL COMPUTING

popular suggestions during the event. For the enhanced search function, the AEP provided a list of related keywords and a metric to quantify the relevance to the search request. Figure 7 shows a heat map of related terms in a social network generated from data output by the AEP on race day (6 July 2014). The darker the episode of the related keyword, the higher its relevance to the keyword “silverstone.” This illustrates how hot topics related to live events shared on social me dia evolve over time as influenced by how the event develops. For instance, “ #pinkforpapa,” a story about the Brit ish race car driver Jenson Button and his father, was a trending topic prior to Figure 6. AEP experiments during the 2014 Winter Olympics. the race (http://en.espn.co.uk/f1/motor sport/story/164969.html). This demon strates how the AEP identifies social discussions on background stories of an found the tweet interesting and rele audio visual content (not only still im event. The hotness of the term “kimi” in vant to the event by clicking a green (in ages) for personal archiving or social the heat map reflects our observations teresting and relevant) or red (not inter sharing, especially when a group of that the crash on the first lap (by driver esting or relevant) button (see Figure 6). friends or family travel together. Kimi Raikkonen) quickly became a The main keyword to start the profiler The videos uploaded by the user trending topic on social media but was “#os2014,” the hashtag promoted community covered the entire trip to didn’t last later in the race when topics by the television broadcaster NOS. A Silverstone. We noticed that only a changed to specific battles and the win list of keywords generated by the event small portion of the videos were about ner of the race (“hamilton,” “#teamlh,” profiler was updated every 5 minutes. the actual race, while most others were and so on). The related keywords were We found that the AEP retrieved sig related to other parts of the Silverstone effective in helping participants dis nificantly more likeable and interest experience, including traveling, camp cover and create personal stories in the ing tweets compared with results from ing, and auxiliary entertainment. STEER system. following a single keyword, without in To improve the experience for us troducing too much noise. ers of the STEER system, we analyzed We also found that relevant tweets social media and leveraged a deep un hrough our research and ex aren’t necessarily interesting, but in derstanding of community activities, periments, we witnessed the teresting tweets are usually relevant. which proved to be effective in improv Tunique role of user creativity ing video annotation and enhancing in the development cycle of creating 2014 Silverstone the search function during the exper and sharing engaging social stories. Formula 1 Grand Prix iments. Using the AEP to recognize Furthermore, we found that social A third experiment was held during trending events ideally solves the “cold context derived from social media, the Silverstone Formula 1 British Grand start” problem in content recommen locationbased services, and emerging Prix racing event on 6 July 2014 in Eng dation, when a system doesn’t have mobile technologies can also greatly land. We selected two groups of mem any prior knowledge of the user’s pref improve the story capturing and au bers of the public as participants. Both erences. Given user location, STEER is thoring process. Although automated groups used the storytelling applica able to suggest a number of popular social storytelling systems are becom tions of the STEER system extensively user stories related to socially trend ing more prevalent, humanlevel in to capture the highlights of their Sil ing keywords nearby. The main key sights, perception, and creativity can’t verstone experience. The amount of word used as input for the AEP in this be replaced by a computer algorithm. data submitted by the groups suggests case was “silverstone.” Event and par Rather, humans will continue to be an that participants of live events are very ticipant names (such as “#britishgp” active and catalytic force in eliciting likely to capture a large amount of and “hamilton”) were among the most immersive community experiences.

14 ComputingEdge August 2016 96 COMPUTER WWW.COMPUTER.ORG/COMPUTER SOCIAL COMPUTING

popular suggestions during the event. #f1 For the enhanced search function, the #britishgp AEP provided a list of related keywords and a metric to quantify the relevance hamilton to the search request. #silverstone Figure 7 shows a heat map of related race terms in a social network generated from data output by the AEP on race lewis day (6 July 2014). The darker the episode #pinkforpapa of the related keyword, the higher its #formula1 relevance to the keyword “silverstone.” This illustrates how hot topics related weekend Keywords to live events shared on social me #teamlh dia evolve over time as influenced by how the event develops. For instance, wimbledon “ #pinkforpapa,” a story about the Brit #skyf1 ish race car driver Jenson Button and kimi Related his father, was a trending topic prior to Candidate Figure 6. AEP experiments during the 2014 Winter Olympics. the race (http://en.espn.co.uk/f1/motor wins Race time sport/story/164969.html). This demon rosberg strates how the AEP identifies social discussions on background stories of an 3:00 am 6:00 am 9:00 am 12:00 pm 3:00 pm 6:00 pm 9:00 pm Mon 7 July found the tweet interesting and rele audio visual content (not only still im event. The hotness of the term “kimi” in Time vant to the event by clicking a green (in ages) for personal archiving or social the heat map reflects our observations Figure 7. Heat map of related social media terms on the Silverstone Formula 1 race day (6 July 2014). teresting and relevant) or red (not inter sharing, especially when a group of that the crash on the first lap (by driver esting or relevant) button (see Figure 6). friends or family travel together. Kimi Raikkonen) quickly became a The main keyword to start the profiler The videos uploaded by the user trending topic on social media but With the rise of network virtual- Watching to Community Story Tell- was “#os2014,” the hashtag promoted community covered the entire trip to didn’t last later in the race when topics ization, the Internet of Things, and ing,” IEEE Comm. Magazine, OMAR NIAMUT is a senior research by the television broadcaster NOS. A Silverstone. We noticed that only a changed to specific battles and the win emerging wireless technologies, we vol. 51, no. 8, 2013, pp. 112–119. scientist at TNO. Contact him at list of keywords generated by the event small portion of the videos were about ner of the race (“hamilton,” “#teamlh,” envision more synergic actions among 6. M. Mu et al., “‘Let’s Share a Story’: [email protected]. profiler was updated every 5 minutes. the actual race, while most others were and so on). The related keywords were social media, networking, and ma- Socially Enhanced Multimedia Story- We found that the AEP retrieved sig related to other parts of the Silverstone effective in helping participants dis chine learning. telling,” IEEE MultiMedia, vol. 22, MU MU is an assistant professor nificantly more likeable and interest experience, including traveling, camp cover and create personal stories in the no. 3, 2015, pp. 54–65. in the Department of Computing ing tweets compared with results from ing, and auxiliary entertainment. STEER system. REFERENCES 7. H. Stokking et al., “Augmenting a TV and Immersive Technologies at the following a single keyword, without in To improve the experience for us 1. F. Chen et al., “Cloud-Assisted Broadcast with Synchronised User University of Northampton. Contact troducing too much noise. ers of the STEER system, we analyzed Live Streaming for Crowdsourced Generated Video and Relevant Social him at [email protected]. We also found that relevant tweets social media and leveraged a deep un hrough our research and ex Multimedia Content,” IEEE Trans. Network Content,” Proc. ACM Int’l aren’t necessarily interesting, but in derstanding of community activities, periments, we witnessed the Multi media, vol. 17, no. 9, 2015, Conf. Interactive Experiences for TV and SPYROS DENAZIS is an associate teresting tweets are usually relevant. which proved to be effective in improv unique role of user creativity pp. 1471–1483. Online Video (TVX 14), 2014; http:// professor in the Department of T Electrical and Computer Engineering ing video annotation and enhancing in the development cycle of creating 2. H. Wang, F. Wang, and J. Liu, “Tor- repository.tudelft.nl/view/tno/uuid 2014 Silverstone the search function during the exper and sharing engaging social stories. rents on Twitter: Explore Long-Term %3Aa4ad12df-2e12-488d-9a79-cf19 at the University of Patras. Contact Formula 1 Grand Prix iments. Using the AEP to recognize Furthermore, we found that social Social Relationships in Peer-to-Peer c72a9951. him at [email protected]. A third experiment was held during trending events ideally solves the “cold context derived from social media, Systems,” IEEE Trans. Network and 8. M. Efthymiopoulou et al., “Scalable the Silverstone Formula 1 British Grand start” problem in content recommen locationbased services, and emerging Service Management, vol. 10, no. 1, Playback Rate Control in P2P Live NICHOLAS RACE is a reader in Prix racing event on 6 July 2014 in Eng dation, when a system doesn’t have mobile technologies can also greatly 2012, pp. 95–104. Streaming Systems,” Peer-to Peer the School of Computing and land. We selected two groups of mem any prior knowledge of the user’s pref improve the story capturing and au 3. Z. Wang et al., “CPCDN: Content De- Networking and Applications, Communications at Lancaster bers of the public as participants. Both erences. Given user location, STEER is thoring process. Although automated livery Powered by Context and User Springer, 2015; doi: 10.1007/s12083 University. Contact him at n.race@ groups used the storytelling applica able to suggest a number of popular social storytelling systems are becom Intelligence,” IEEE Trans. Multimedia, -015-0403-6. lancaster.ac.uk. tions of the STEER system extensively user stories related to socially trend ing more prevalent, humanlevel in vol. 17, no. 1, 2014, pp. 92–103. 9. R. Kaptein et al., “Retrieving Rele- to capture the highlights of their Sil ing keywords nearby. The main key sights, perception, and creativity can’t 4. J. Wu et al., “Context-aware Network- vant and Interesting Tweets During verstone experience. The amount of word used as input for the AEP in this be replaced by a computer algorithm. ing and Communications: Part 1,” Live Television Broadcasts,” Current Selected CS articles and data submitted by the groups suggests case was “silverstone.” Event and par Rather, humans will continue to be an IEEE Comm. Magazine, vol. 52, no. 6, Trends in Web Eng., LNCS 9396, F. This articlecolumns originally are also available appeared for in that participants of live events are very ticipant names (such as “#britishgp” active and catalytic force in eliciting 2014, pp. 14–15. Daniel and O. Diaz, eds., Springer, Computerfree at, vol. http://ComputingNow 49, no. 5, 2016. likely to capture a large amount of and “hamilton”) were among the most immersive community experiences. 5. M. Mu et al., “Storisphere: From TV 2015, pp. 175–185. .computer.org.

www.computer.org/computingedge 15 96 COMPUTER WWW.COMPUTER.ORG/COMPUTER MAY 2016 97 Natural Web Interfaces Editor: Munindar P. Singh • [email protected] Engineering Privacy in Social Applications

Pradeep K. Murukannaiah, Nirav Ajmeri, and Munindar P. Singh • North Carolina State University

Experience with a social application depends crucially upon how it supports or interferes with the users’ social expectations. Because privacy is central to the user’s experience, the authors introduce Danio, a methodology based on modeling users’ expectations in various contexts.

illiam Prosser, in his classic paper, and the secondary users are callers and those within describes how privacy involves diverse earshot. The ringer manager helps the owner set a aspects centered on the idea of being let ringer mode (loud, silent, or vibrate) for incoming W 1 alone (p. 389). We view privacy as crucial to the phone calls. A rigid design yields poor privacy and user experience (UX) in a sociotechnical setting, experience: The phone might ring loudly when the wherein users engage socially through informa- owner is in an important meeting (causing a nui- tion technology. In previous work, we addressed sance) or stay silent even when the owner’s spouse privacy by limiting information disclosure to calls in an emergency (losing value). location-based (and not necessarily social) appli- Traditionally, UX design concentrates on pri- cations by computing context at an abstract level mary users and disregards secondary users.3 This that enhances usability but hides details.2 attitude can lead to suboptimal experiences for Here, we approach privacy from the stand- both primary and secondary users, specifically, point of engineering social applications, wherein because privacy presupposes interaction between interactions among users are central and thus users. privacy matters for more than just disclosure. Intrusion is a prominent aspect of privacy in Specifically, we investigate how to develop appli- the ringer scenario: Does the caller intrude upon cations that deal with two foundational aspects the callee and does the callee intrude upon people of privacy.1 Intrusion into someone’s solitude, nearby (by taking a call or by letting a phone which originally meant physical intrusion into a ring)? The UX for all concerned parties depends person’s space, we also take to include making upon whether the phone rings: the caller could be a noise or otherwise interrupting someone’s life. stymied by a phone set on silent and the privacy Disapprobation of someone’s peers would mean of the other users might be violated otherwise. loss of face, which can arise not only with a per- Also, improper ringer settings expose another son’s inner circle but also with strangers looking privacy risk — that of disapprobation, causing askance. Avoiding intrusion and disapprobation the owner embarrassment. Imagine if your phone is overshadowed in computing research by con- went off during a classical concert! cerns of information leakage — arguably, information leakage involves confidentiality more Social Expectations than privacy per se. Producing privacy-enhancing controls is non- A social application caters to multiple users: trivial. Setting a fixed ringer mode for all incom- primary users, who directly interact with it, and ing calls (as is common today) ignores secondary secondary users, who might not directly interact users, but asks the owner to anticipate contexts with the application but are affected by it. The and secondary users. But setting appropriate ringer lowly ringer manager on a cell phone is a social policies is too complex to be viable. Accordingly, application: its primary user is the phone’s owner we adopt the idea of modeling social interactions.

72 Published by the IEEE Computer Society 1089-7801/16/$33.00 © 2016 IEEE IEEE INTERNET COMPUTING 16 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE Natural Web Interfaces Engineering Privacy in Social Applications Editor: Munindar P. Singh • [email protected] Engineering Privacy Context

Primary Secondary in Social Applications Performs Action Affects user user In uences Species Pradeep K. Murukannaiah, Nirav Ajmeri, and Munindar P. Singh • North Carolina State University Application User Renes Satises or violates Results in Renes designer experience Experience with a social application depends crucially upon how it supports or interferes with the users’ social expectations. Because privacy is central to the user’s experience, the authors introduce Danio, a methodology based on Species Norm In uences Sanction In uences modeling users’ expectations in various contexts. Social expectation illiam Prosser, in his classic paper, and the secondary users are callers and those within describes how privacy involves diverse earshot. The ringer manager helps the owner set a Figure 1. A conceptual model showing how a primary user’s action within a certain context affects the secondary user, aspects centered on the idea of being let ringer mode (loud, silent, or vibrate) for incoming and how contexts and expectations influence the user experience. A traditional design emphasizes only the primary users’ W 1 alone (p. 389). We view privacy as crucial to the phone calls. A rigid design yields poor privacy and actions and contexts, but this model additionally captures how a primary user’s action might affect another (secondary) user. user experience (UX) in a sociotechnical setting, experience: The phone might ring loudly when the wherein users engage socially through informa- owner is in an important meeting (causing a nui- tion technology. In previous work, we addressed sance) or stay silent even when the owner’s spouse Katja Battarbee4 motivates co-experi- tions fit into this sphere. Spiekermann sizes contextual design — an important privacy by limiting information disclosure to calls in an emergency (losing value). ence as a seamless blend of UX and and Cranor identify the key engineer- part of the well-known user-centered location-based (and not necessarily social) appli- Traditionally, UX design concentrates on pri- social interactions. ing challenge in this sphere as provid- design process. Whereas traditional cations by computing context at an abstract level mary users and disregards secondary users.3 This However, social interaction is a ing individuals’ control over access to design emphasizes the primary user’s that enhances usability but hides details.2 attitude can lead to suboptimal experiences for loose concept. We propose instead to themselves. In contrast, we argue that actions and contexts, this model addi- Here, we approach privacy from the stand- both primary and secondary users, specifically, capture users’ social expectations of a key challenge in this sphere, perhaps tionally captures how a primary user’s point of engineering social applications, wherein because privacy presupposes interaction between others as central to delivering an opti- more important than control, is to action might affect another (hence, sec- interactions among users are central and thus users. mal experience. For example, the ringer engineer solutions where both senders ondary) user. A consequence of social- privacy matters for more than just disclosure. Intrusion is a prominent aspect of privacy in manager’s UX depends on whether and recipients are accountable to each ity is that the context helps determine Specifically, we investigate how to develop appli- the ringer scenario: Does the caller intrude upon the phone owner meets their spouse’s other and thus vested in enhancing whether a user is primary or secondary. cations that deal with two foundational aspects the callee and does the callee intrude upon people expectation by answering important each other’s privacy. The lower part of this model cap- of privacy.1 Intrusion into someone’s solitude, nearby (by taking a call or by letting a phone calls, which in turn might depend on We adopt two interrelated compu- tures social expectations. A primary which originally meant physical intrusion into a ring)? The UX for all concerned parties depends whether the ringer manager lets the tational abstractions to capture social user’s action, within a context, might person’s space, we also take to include making upon whether the phone rings: the caller could be phone ring. Thus, we seek to systemati- expectations. First, a norm character- satisfy or violate a norm directed a noise or otherwise interrupting someone’s life. stymied by a phone set on silent and the privacy cally incorporate social expectations in izes a user’s expected (“normal”) behav- toward a secondary user, leading the Disapprobation of someone’s peers would mean of the other users might be violated otherwise. the UX as geared toward privacy. ior from the perspective of another latter to sanction (reward or punish) loss of face, which can arise not only with a per- Also, improper ringer settings expose another Broadly, looking beyond privacy, user.6 Second, a sanction characterizes the primary user. son’s inner circle but also with strangers looking privacy risk — that of disapprobation, causing social expectations and their influences a user’s response to another user’s satis- Based on this information, we askance. Avoiding intrusion and disapprobation the owner embarrassment. Imagine if your phone are abundant in real-life interactions. faction or violation of a norm.7,8 Norms understand — in a manner largely is overshadowed in computing research by con- went off during a classical concert! Expectations arise both explicitly (as in and sanctions arising in an applica- unique to our approach — that a norm cerns of information leakage — arguably, infor- text-based interactions) and implicitly tion’s context can be computationally is a directed normative relationship. mation leakage involves confidentiality more Social Expectations (as via gestures). Yet, current UX tech- represented and reasoned about. (For the remainder of this article, we than privacy per se. Producing privacy-enhancing controls is non- niques largely disregard expectations. Figure 1 captures a model in which use “norm” in this sense.) A norm A social application caters to multiple users: trivial. Setting a fixed ringer mode for all incom- How can we computationally represent a user’s actions in a usage context helps capture application require- primary users, who directly interact with it, and ing calls (as is common today) ignores secondary and reason about expectations? influence social expectations, and both ments in terms of what one stake- secondary users, who might not directly interact users, but asks the owner to anticipate contexts Sarah Spiekermann and Lorrie Faith contexts and expectations influence holder expects from another.6 A norm with the application but are affected by it. The and secondary users. But setting appropriate ringer Cranor5 describe a “joint sphere” of the user experience. The application is directed from a subject to an object lowly ringer manager on a cell phone is a social policies is too complex to be viable. Accordingly, privacy responsibilities involving data designer specifies both contexts and and is constructed as a conditional application: its primary user is the phone’s owner we adopt the idea of modeling social interactions. senders and recipients. Social expecta- social expectations. This model empha- relationship involving an antecedent

Actor modeling 1. Identify actors (primary and secondary users; system components) 2. Identify each actor’s goals and plans pertinent to the application

Context modeling 3. Identify contexts in which actors pursue goals and execute plans

4. Identify applicable norms between actors in each context Social expectation 5. Identify sanctions for satisfying and violating norms modeling 6. Identify and resolve potential con icts between norms

Figure 2. Danio summarized. The methodology helps systematically incorporate social expectations into application design.

(which brings the norm in force) and Understanding an application in methodology, called Danio (after the a consequent (which brings the norm terms of social expectations yields fish), for systematically incorporat- to satisfaction). This representation key payoffs in terms of privacy. ing social expectations into applica- yields clarity on who’s accountable to tion design by extending Tropos9 and whom. A norm follows this template: • Personalization. Decentralized enact- Xipho.10 As Figure 2 shows, Danio’s ment facilitates modeling software key phases are modeling actors, con- N(SUBJECT, OBJECT, antecedent, functionality — independently of texts, and expectations. consequent). implementation — that respects users’ (subjective) privacy. Modeling Actors Each stakeholder is autonomous, • Disclosure and control. Explicit Actor modeling identifies prospective meaning it can violate any norm. social expectations advise a user users and their requirements. Danio However, an application — and broadly on precisely what information to adopts the following modeling con- speaking, society — operates under the provide others, thereby avoiding structs from Tropos.9 assumption that most people respect unnecessary disclosure. An actor is a user role or software these norms. How can we coordinate • Accountability. Each norm deter- agent. actions toward a predefined or an mines who’s accountable to whom, A goal is an actor’s strategic inter- emergent social order? A sanction7,8 in what context, and what the con- est that the application would serve. specifies the consequences its subject comitant sanctions are, thereby Goals can decompose into subgoals. faces from its object for satisfying or helping to operationalize the expec- A plan abstracts actions to satisfy violating a norm, thereby promoting tations to promote privacy. a goal. compliance. We write a sanction as In the ringer scenario, the actors are Our approach presumes represent- the CALLEE (owner), CALLER (owner’s S(SUBJECT, OBJECT, antecedent, ing context and expectations. At one spouse), CALLEE’S NEIGHBOR (meet- consequent). extreme, a designer could produce a ing participants), and RINGER (the soft- complete specification; at the other, ware). The CALLEE’s goals include don’t A positive sanction rewards or encour- contexts and expectations could emerge disturb neighbors. The RINGER’s plans ages compliance and a negative sanc- at runtime through active usage. We include set silent mode. tion penalizes or discourages violation. adopt a pragmatic middle-ground solu- Sanction types include the following: tion: the designer formulates an incom- Modeling Context autonomic, where the consequence plete specification, which users refine as We identify contexts in which actors act of the norm violated or satisfied is they interact. and interact. However, context is an all- in itself a sanction; material, often encompassing notion. Which contexts financial; social, essentially, affecting Danio are relevant to a given application? a reputation; and psychological, such Keeping all of these criteria in mind, We understand context as inher- as guilt.8 here we propose a software engineering ently related to the actors’ goals and

74 www.computer.org/internet/ IEEE INTERNET COMPUTING 18 ComputingEdge August 2016 Natural Web Interfaces Engineering Privacy in Social Applications

plans,10 which provides a systematic it’s satisfied or violated. The sanc- Preliminary Evaluation basis for eliciting relevant contexts. tion’s SUBJECT and OBJECT are the We conducted a developer study to Actor modeling 1. Identify actors (primary and secondary users; system components) Specifically, situations such as the corresponding norm’s OBJECT and evaluate our methodology. We asked 2. Identify each actor’s goals and plans pertinent to the application following provide grounds for elicit- SUBJECT, respectively. Its antecedent subjects (34 graduate computer science ing contexts. captures the status of the norm and students, working solo) to develop a its consequent the sanctioning action. ringer manager. We split the subjects Context modeling 3. Identify contexts in which actors pursue goals and execute plans Conflicting goals. In this scenario, an The following are examples of into two groups: control (n = 16) and actor can’t satisfy all of the goals. For norms and sanctions in our example. Danio (n = 18), providing identical example, the CALLEE’s goals answer application requirements to each, but 4. Identify applicable norms between actors in each context important calls and don’t disturb • The CALLEE is committed to his providing our methodology (treat- Social expectation 5. Identify sanctions for satisfying and violating norms modeling neighbors conflict. This choice might spouse for answering the spouse’s ment) only to the Danio subjects. All 6. Identify and resolve potential con icts between norms depend upon contextual elements calls: subjects recorded their development such as locale as well as the CALLEE’s C1 (CALLER, CALLEE, relationship time during the study and completed relationships with the CALLER and = spouse ∧ call, answer call). a post-survey. Figure 2. Danio summarized. The methodology helps systematically incorporate social expectations into application design. NEIGHBOR. • The CALLEE is prohibited by his Development Time Multiple plans to satisfy the same coworkers from answering calls in Figure 3 shows boxplots of times goal. For example, the RINGER’s plans meetings: expended by subjects during different (which brings the norm in force) and Understanding an application in methodology, called Danio (after the set silent mode and set vibrate mode P1 (CALLER, NEIGHBOR, relation- development phases. The diamond a consequent (which brings the norm terms of social expectations yields fish), for systematically incorporat- can each satisfy the goal don’t disturb ship = coworker ∧ place = meet- dots indicate mean values and the to satisfaction). This representation key payoffs in terms of privacy. ing social expectations into applica- neighbors. The RINGER might choose ing ∧ call, answer call). other dots indicate outliers. yields clarity on who’s accountable to tion design by extending Tropos9 and at most one of these plans at runtime. We observed that the Danio subjects 10 whom. A norm follows this template: • Personalization. Decentralized enact- Xipho. As Figure 2 shows, Danio’s Its choice could depend upon the CAL- • Satisfaction of C1 increases the took slightly longer than the control ment facilitates modeling software key phases are modeling actors, con- LEE’s ambiance. trust (positive sanction) of the subjects to understand requirements N(SUBJECT, OBJECT, antecedent, functionality — independently of texts, and expectations. Such scenarios help tailor the spouse toward the CALLEE: and prepare a specification. However, consequent). implementation — that respects generic context model to an applica- S1 (CALLEE, CALLER, C1 = satis- Danio subjects, on average, spent 17.8 users’ (subjective) privacy. Modeling Actors tion-specific model. For example, the fied, increase trust). and 11.4 percent less time than control Each stakeholder is autonomous, • Disclosure and control. Explicit Actor modeling identifies prospective RINGER’s context model can include subjects in implementing and testing meaning it can violate any norm. social expectations advise a user users and their requirements. Danio the CALLEE’s relationships with the • Violation of P1 yields a bad repu- the application, respectively. We posit However, an application — and broadly on precisely what information to adopts the following modeling con- CALLER and NEIGHBOR. Some aspects tation (negative sanction) for the that Danio’s systematization of incor- speaking, society — operates under the provide others, thereby avoiding structs from Tropos.9 of context are user-specific and can be CALLEE among his coworkers: porating context, which costs extra assumption that most people respect unnecessary disclosure. An actor is a user role or software elicited from users at runtime via a con- C2 (NEIGHBOR, CALLER, P1 = time early, pays off during implemen- these norms. How can we coordinate • Accountability. Each norm deter- agent. textual middleware.2 violated, bad-mouth). tation and testing. actions toward a predefined or an mines who’s accountable to whom, A goal is an actor’s strategic inter- emergent social order? A sanction7,8 in what context, and what the con- est that the application would serve. Modeling Social Expectations Additional norms can be defined, Post-Survey Data specifies the consequences its subject comitant sanctions are, thereby Goals can decompose into subgoals. We model social expectations between limited only by the social context. For We asked control subjects the extent faces from its object for satisfying or helping to operationalize the expec- A plan abstracts actions to satisfy actors. Whereas Tropos and Xipho model example, we might state that a univer- to which a methodology would help violating a norm, thereby promoting tations to promote privacy. a goal. broad-brush dependencies between sity librarian has the power to declare them in developing a social applica- compliance. We write a sanction as In the ringer scenario, the actors are actors, we derive a detailed specification reading rooms as quiet or as places tion. We used a Likert scale of one (not Our approach presumes represent- the CALLEE (owner), CALLER (owner’s of expectations in terms of norms and where discussions are allowed. helpful) to seven (extremely helpful) S(SUBJECT, OBJECT, antecedent, ing context and expectations. At one spouse), CALLEE’S NEIGHBOR (meet- sanctions involving actors, actions, and Norms might conflict, as when the for each question; x are mean values. consequent). extreme, a designer could produce a ing participants), and RINGER (the soft- contexts. phone owner receives a call from their The subjects responded as follows: complete specification; at the other, ware). The CALLEE’s goals include don’t spouse during a meeting with cowork- that a methodology would be helpful A positive sanction rewards or encour- contexts and expectations could emerge disturb neighbors. The RINGER’s plans Norms. For each actor, we iden- ers. In this case, antecedents of both C1 for understanding requirements (x = ages compliance and a negative sanc- at runtime through active usage. We include set silent mode. tify norms where the actor is the and P1 hold, but satisfying one norm 5); implementation (x = 4.5); testing tion penalizes or discourages violation. adopt a pragmatic middle-ground solu- SUBJECT and another actor is the means violating the other. A designer (x = 5.5); and specifications (x = 3). Sanction types include the following: tion: the designer formulates an incom- Modeling Context OBJECT. The context in which the must identify such conflicts and elicit Surprisingly, control subjects felt autonomic, where the consequence plete specification, which users refine as We identify contexts in which actors act norm applies can be captured in its additional contextual information to that a methodology (which we didn’t of the norm violated or satisfied is they interact. and interact. However, context is an all- antecedent and the expected behavior prioritize among the conflicting norms. provide to them) would be least helpful in itself a sanction; material, often encompassing notion. Which contexts in its consequent. Our conflict-resolution methodology,11 for application specifications, whereas financial; social, essentially, affecting Danio are relevant to a given application? based on an analysis of competing Danio subjects felt a methodology a reputation; and psychological, such Keeping all of these criteria in mind, We understand context as inher- Sanctions. For each norm, we iden- hypotheses, can guide a designer in (which we did provide to them) was as guilt.8 here we propose a software engineering ently related to the actors’ goals and tify the sanctions that apply when prioritizing the alternatives. the most helpful for specifications

74 www.computer.org/internet/ IEEE INTERNET COMPUTING MaRCh/aPRIl 2016 75 www.computer.org/computingedge 19 Natural Web Interfaces

Socio-Technical Systems,” The Knowledge Control Control Eng. Rev., vol. 31, no. 2 2016, pp. 1–24. 9. P. Bresciani et al., “Tropos: An Agent- Danio Danio Oriented Software Development Methodol- 0 500 1,000 1,500 100 200 300 (a)(b) ogy,” Autonomous Agents and Multi-Agent Systems, vol. 8, no. 3, 2004, pp. 203–236. 10. P.K. Murukannaiah and M.P. Singh, “Xipho: Control Control Extending Tropos to Engineer Context-Aware Danio Danio Personal Agents,” Proc. Int’l Conf. Autono- 0 500 1,000 1,500 0 200 400 600 800 (c) (d) mous Agents and Multi-Agent Systems, 2014, pp. 309–316. 11. P.K. Murukannaiah et al., “Resolving Goal Figure 3. Time expended (in minutes) by subjects in developing the ringer Conflicts via Argumentation-Based Analy- manager. Time to (a) understand requirements, (b) prepare the specification, sis of Competing Hypotheses,” Proc. 23rd (c) implement, and (d) test. IEEE Int’l Requirements Eng. Conf., 2015, pp. 156–165.

(x = 5). The application specifica- tradeoffs is an important challenge for tions can be quite valuable in testing, future research. Pradeep K. Murukannaiah is a computer sci- and subsequently, for maintenance ence PhD candidate at North Carolina State (although our study didn’t include Acknowledgment University. His research interests include maintenance). We thank the US Department of Defense for software engineering, social computing, support through the Science of Security Lablet. and context-aware systems. Murukannaiah has an MS in computer science from North ncorporating social expectations References Carolina State University. Contact him at I enables an application to deliver 1. W.L. Prosser, “Privacy,” California Law [email protected]. a privacy-preserving experience by Rev., vol. 48, no. 3, 1960, pp. 383–423. promoting personalization, disclosure 2. P.K. Murukannaiah and M.P. Singh, Nirav Ajmeri is a computer science PhD stu- and control, and accountability. Danio “Platys: An Active Learning Framework dent at North Carolina State University. His extends well-known design and engi- for Place-Aware Application Development research interests include software engineering techniques to engineer privacy and Its Evaluation,” ACM Trans. Software neering and multiagent systems. Ajmeri into social applications by incorporat- Eng. and Methodology, vol. 24, no. 3, 2015, has a BE in computer engineering from ing social expectations. Our preliminary pp. 1–33. Sardar Vallabhbhai Patel Institute of Tech- evaluation shows the merits of Danio 3. O.A. Alsos and D. Svanæs, “Designing for nology, Gujarat University. Contact him at and sets the stage for more extensive the Secondary User Experience,” P. Campos [email protected]. evaluations. et al., eds., Proc. Human-Computer Interac- An interesting direction is extend- tion, LNCS 6949, Springer, 2011, pp. 84–91. Munindar P. Singh is a computer science profes- ing Danio to tackle cases where user 4. K. Battarbee, “Defining Co-Experience,” sor at North Carolina State University. His interactions are motivated by subtle Proc. Int’l Conf. Designing Pleasurable research interests include the engineering tradeoffs between privacy and social Products and Interfaces, 2003, pp. 109–113. and governance of sociotechnical systems. utility. For example, a callee might 5. S. Spiekermann and L.F. Cranor, “Engineer- Singh is an IEEE Fellow, a former Editor-in- accept a call to gloat to his neighbors ing Privacy,” IEEE Trans. Software Eng., Chief of IEEE Internet Computing, and the about the caller having called him or vol. 35, no. 1, 2009, pp. 67–82. current Editor-in-Chief of ACM Transac- to the caller about a meeting with the 6. M.P. Singh, “Norms as a Basis for Govern- tions on Internet Technology. Contact him at neighbors. ing Sociotechnical Systems,” ACM Trans. [email protected]. One contribution of Danio is to Intelligent Systems and Technology, vol. 5, show how to engineer privacy in social no. 1, 2013, pp. 21:1–21:23. applications in a way that accommo- 7. P. Pasquier, R.A. Flores, and B. Chaib-draa, dates aspects of privacy that are often “An Ontology of Social Control Tools,” Proc. de-emphasized. Synthesizing these Int’l Joint Conf. Autonomous Agents and This article originally appeared in diverse aspects of privacy and providing Multiagent Systems, 2006, pp. 1369–1371. Selected CS articles and columns IEEE Internet Computing, vol. 20, a context-sensitive way to engi- 8. L.G. Nardin et al., “Classifying Sanctions are also available for free at http:// no. 2, 2016. neer privacy with respect to diverse and Modelling a Sanctioning Process for ComputingNow.computer.org.

76 www.computer.org/internet/ IEEE INTERNET COMPUTING 20 ComputingEdge August 2016 IEEE Computer Society 2016 Call for MAJOR AWARD NOMINATIONS Help Recognize Computing’s Most Prestigious

IEEE Computer Society awards recognize outstanding achievements and highlight significant contributors in the teaching and R&D computing communities. All members of the profession are invited to nominate individuals who they consider most eligible to receive international recognition of an appropriate society award.

Computer Entrepreneur Award Computer Pioneer Award Sterling Silver Goblet Silver Medal Vision and leadership resulting in the growth of Pioneering concepts and development of the some segment of the computer industry. computer field.

Technical Achievement Award W. Wallace McDowell Award Certificate/$2,000 Certificate/$2,000 Contributions to computer science or computer Recent theoretical, design, educational, technology. practical, or other tangible innovative contributions. Harry H. Goode Memorial Award Bronze Medal/$2,000 Taylor L. Booth Award Information sciences, including seminal ideas, Bronze Medal/$5,000 algorithms, computing directions, and concepts. Contributions to computer science and engineering education. Hans Karlsson Award Plaque/$2,000 Computer Science & Engineering Team leadership and achievement through Undergraduate Teaching Award collaboration in computing standards. Plaque/$2,000 Recognizes outstanding contributions to Richard E. Merwin Distinguished Service undergraduate education. Award Bronze Medal/$5,000 IEEE-CS/Software Engineering Institute Outstanding volunteer service to the profession Watts S. Humphrey Software Process at large, including service to the IEEE Computer Achievement Award Society. (Joint award by CS/SEI) Plaque/$1,500 Harlan D. Mills Award Software professionals or teams responsible for Plaque/$3,000 an improvement to their organization’s ability to Contributions to the practice of software create and evolve software-dependent systems. engineering through the application of sound theory.

Deadline: 15 October 2016 Nomination Site: awards.computer.org For more information visit: www.computer.org/awards SOCIAL COMPUTING

How Socially Aware Are Social Media Privacy Controls?

Gaurav Misra and Jose M. Such, Lancaster University Despite social media’s unques- tionable success, privacy concerns Social media sites are key mediators of online have been increasing over the past few years. Sharing personal infor- communication. Yet the privacy controls for mation with unintended audiences these sites are not fully socially aware, even (commonly referred to as the “insider threat”) is of particular con- when privacy management is known to be cern. This is especially problematic with sites that treat all of a user’s fundamental to successful social relationships. “friends” in the same manner—for example, without di erentiating between close friends and profes- he use of social media sites such as Facebook, sional colleagues. Along with other variables such as lo- Twitter, Google+, LinkedIn, and Pinterest has cation, topic, and time, the social context of a particular grown massively over the last decade. Social disclosure forms a very signi cant part of that disclo- media users experience many bene ts, such as sure’s overall context. For example, if a user is posting Testablishing, developing, and maintaining social rela- about an event in his personal life, he might choose to tionships through shared personal information. Indeed, deny access to his colleagues. The context here is de- this plays a crucial role in managing what is known as so- ned by the topic of the disclosure as well as the type cial capital. Social media sites also enable users to build of social relationship the user has with the audience. and participate in communities and networks of people This helps the user determine the appropriate audience who share similar interests. Even governments and poli- for that disclosure, ensuring maintenance of contextual ticians around the world use social media to engage with integrity and enabling regulation of dynamic social citizens. For example, the UK government encourages boundaries. Thus, acknowledging and accommodating civil servants to use social media and provides guide- such social contexts is imperative to safeguard the pri- lines for doing so, and social media arguably played an vacy of social media users. This raises the question: How important role in the reelection of US President Barack socially aware are current mainstream social media pri- Obama in . vacy controls?

22 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 96 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE EDITOR CHRISTIAN TIMMERER Alpen-Adria-Universität Klagenfurt; SOCIAL COMPUTING [email protected]

14 Xing 13 Wayn 12 StumbleUpon 11 Skyrock 10 Pintrest 9 Ning 8 Myspace 7 Meetup Twitter 6 Gaia Online Path

No. of social media sites No. 5 Friendster LiveJournal 4 How Socially Aware Foursquare Tribe Tagged Google+ 3 Flixter LinkedIn Last.fm Fotolog 2 Care2 Fotki hi5 Flickr 1 Are Social Media Buzznet Bebo Geni Experience Project Facebook 0 Binary groups Prede ned groups Prede ned groups User-de ned groups Computer-supported + individuals + individuals groups Privacy Controls? Groups Figure 1. Top 30 social media sites as of January 2015. Sites are classifi ed according to the type of social context modeling mecha- Gaurav Misra and Jose M. Such, Lancaster University nism they provide, and are listed in alphabetical order for each level. Despite social media’s unques- tionable success, privacy concerns Social media sites are key mediators of online have been increasing over the past communication. Yet the privacy controls for few years. Sharing personal infor- CLASSIFICATION OF various access control mechanisms. one or two individuals—the post will mation with unintended audiences MAINSTREAM SOCIAL In our evaluation, we encountered the be seen by all members of that group. these sites are not fully socially aware, even (commonly referred to as the “in- MEDIA PRIVACY CONTROLS following types of mechanisms. We found four sites with this type of sider threat”) is of particular con- We used Alexa’s tra c rankings to privacy mechanism. when privacy management is known to be cern. This is especially problematic identify the top social media sites Binary groups with sites that treat all of a user’s and clustered them into categories Nearly half () of the sites we eval uated Predefi ned groups + individuals fundamental to successful social relationships. “friends” in the same manner—for o er ing similar support to model were found to have a binary distinction An improvement is found in sites such example, without di erentiating social contexts. We excluded dating between a user’s friends and the rest of as Tagged and hi, which provide the between close friends and profes- sites, online shopping sites, and sites the network. This means that the only option of treating individuals sepa- he use of social media sites such as Facebook, sional colleagues. Along with other variables such as lo- that were too speci c to particular relationship type is “friends,” with no rately from prede ned groups. This Twitter, Google+, LinkedIn, and Pinterest has cation, topic, and time, the social context of a particular populations (for example, Classmates granularity. In these scenarios, a user helps users share content in a more grown massively over the last decade. Social disclosure forms a very signi cant part of that disclo- for US graduates and Naijapals for Ni- is unable to distinguish between two real istic way. For example, a user media users experience many bene ts, such as sure’s overall context. For example, if a user is posting gerians). Figure shows the classi - friends, which doesn’t re ect real-life might have a group for “colleagues” Testablishing, developing, and maintaining social rela- about an event in his personal life, he might choose to cation of the top general- purpose social relationships. but might not want each member to tionships through shared personal information. Indeed, deny access to his colleagues. The context here is de- social networking sites. We evaluated see every post. This type of mecha- this plays a crucial role in managing what is known as so- ned by the topic of the disclosure as well as the type the individual social media infra- Predefi ned groups nism allows for more granularity and cial capital. Social media sites also enable users to build of social relationship the user has with the audience. structures by looking at each site’s Some sites, such as LinkedIn, allow for user control. and participate in communities and networks of people This helps the user determine the appropriate audience privacy policies and settings. In most prede ned groups, which help users who share similar interests. Even governments and poli- for that disclosure, ensuring maintenance of contextual cases, we created user accounts to ex- organize their contacts. Users can cre- User-defi ned groups + individuals ticians around the world use social media to engage with integrity and enabling regulation of dynamic social amine the actual available options for ate groups (for example, “colleagues” Another enhancement is allowing users citizens. For example, the UK government encourages boundaries. Thus, acknowledging and accommodating users, because this information wasn’t or “family”) and then treat these to create their own groups. Google+ and civil servants to use social media and provides guide- such social contexts is imperative to safeguard the pri- directly evident from privacy policies. groups di erently. However, there’s Facebook both provide default groups lines for doing so, and social media arguably played an vacy of social media users. This raises the question: How We created sample user accounts us- no option to make exceptions. For ex- but also allow users to manage their own important role in the reelection of US President Barack socially aware are current mainstream social media pri- ing di erent email addresses and example, a user can’t share a post with circles and lists per individual, more Obama in . vacy controls? perimented with each infrastructure’s everyone in the “family” group except closely re ecting real- life relationships.

Computer-supported groups users to control access to their photos of relationship status, including the We list Facebook separately in this clas- on an iOS platform.7 user’s disclosures as well as reciproca- sification because its privacy controls In our analysis of privacy controls, tions provided by other network mem- provide an enhanced grouping mecha- we noticed a glaring absence of mecha- bers.9 This information would provide nism that other social media sites don’t nisms allowing multiple users to nego- the intelligence for a decision support offer. Facebook’s “smart lists” feature tiate access control decisions for items system, which would help users under- allows users to create lists based on that involve them, such as pictures and stand their relationships and fine-tune features like location and workplace. posts, before these items are posted.8 their information disclosures. A learn- Also, the introduction of the “close Most privacy controls apply the pri- ing mechanism would be required so friends” and “acquaintances” lists ac- vacy preferences of the user making that these systems could understand knowledges the important role of rela- the post. However, if an affected user each user’s behavior and adapt accord- tionship strength. wants the post removed, he or she ingly to provide relevant assistance. must negotiate with the poster using One major concern is that such systems RELATIONSHIP-BASED other means of communication such could become too intrusive, as they ob- ACCESS CONTROLS as email, texts, or private messages. serve and analyze all communications Mainstream social media seems to be Even then, negotiations might happen between users. moving toward more socially aware after the item has been already posted Except for a handful of distrib- privacy controls, which shows recog- and a privacy violation has occurred. uted social media sites that are im- nition of the importance of privacy in Computational mechanisms that plemented by peer-to-peer nodes, the social relationships. However, there’s automate the negotiation process provider mediates communication on still room for improvement to make have been identified as a possible mit- most of these sites. Thus, the provider privacy controls more socially aware. igation. This doesn’t mean that us- can influence the amount and nature Traditional access control ap- ers would lose control; instead, these of communication between individ- proaches such as role-based access mechanisms would suggest possible uals on its platforms. For example, a controls or group-based approaches solutions that all parties would need to recent experiment studied the emo- can’t truly capture the social relation- accept. If users don’t accept the sugges- tional contagion of Facebook status ships among users— relationship-based tion, they need to enter into a manual updates by moderating the amount of access controls (ReBACs) are needed.6 negotiation by other means. The main content received by users from various A ReBAC model uses a different set of challenge is to propose solutions that friends.10 This confirms that the pro- properties to define relationships, re- everyone involved will accept most of vider can modify the content being fed sulting in a more natural disclosure the time. Some mechanisms have been to the user through the network, which of personal information. For example, proposed in this very young discipline, can alter the nature and strength of users can decide that only their close but a more formal study is required relationships that depend on interac- friends can see a specific photo. to understand the conditions under tions between individuals. This also One of the main problems facing which users will make concessions, highlights the fact that institutional ReBACs is usability. The ReBAC model and the variables that make users take privacy—which governs the way so- considers a large number of social stronger or more lenient positions cial media providers handle user data, features, which can help users better during these negotiations. including whether and how they will identify the social context of their dis- Another open challenge is the abil- share data with third parties—can play closures; however, the process of de- ity of privacy controls to adequately a role in social privacy.11 fining relationships and selecting au- support users in making favorable diences needs to be easy. This entails disclosure decisions with regard to improving audience visualization— self-presentation and management of ainstream social media seems the visual cues provided to a user relationship dynamics. Users main- to be moving toward more so- about which friends will have access tain different types of relationships Mcially aware privacy controls. to the information the user posted or is with varying objectives, and the However, these efforts are still a long about to post. Moreover, ReBACs need manner and content of their commu- way from the comprehensive modeling to be computationally lightweight nication with these groups varies ac- of social relationships that would enable and to seamlessly integrate with the cordingly. These nuances must be ade- satisfactory boundary regulation, mim- social media interface, preserving the quately accommodated and supported, icking peoples’ relationships in the off- dynamism of users’ social networking which poses many sociotechnical line world. Current research advances experiences. A recent effort to imple- challenges for future research on the in this area will lead to the development ment a ReBAC mechanism uses at- topic. One particular challenge is the of next- generation privacy controls. tribute-based encryption that allows development of tools that are aware These new privacy controls will, like

24 ComputingEdge August 2016 98 COMPUTER WWW.COMPUTER.ORG/COMPUTER SOCIAL COMPUTING This article originally appeared in Computer, vol. 49, no. 3, 2016.

Computer-supported groups users to control access to their photos of relationship status, including the any new technology, be a double- edged .edu/faculty/rwarner/classes Factors in Computing Systems (CHI 11), We list Facebook separately in this clas- on an iOS platform.7 user’s disclosures as well as reciproca- sword. They will empower users to /internetlaw/2011/materials 2011, pp. 3217–3226. sification because its privacy controls In our analysis of privacy controls, tions provided by other network mem- manage their privacy online in unprec- /nissenbaum_norms.pdf. 9. J.M. Such et al., “Self-Disclosure provide an enhanced grouping mecha- we noticed a glaring absence of mecha- bers.9 This information would provide edented ways, but they could also enable 4. L. Palen and P. Dourish, “Unpacking Decision Making Based on Intimacy nism that other social media sites don’t nisms allowing multiple users to nego- the intelligence for a decision support potentially hazardous, privacy- invasive ‘Privacy’ for a Networked World,” and Privacy,” J. Information Sciences, offer. Facebook’s “smart lists” feature tiate access control decisions for items system, which would help users under- practices, such as more accurate tar- Proc. SIGCHI Conf. Human Factors in vol. 211, 2012, pp. 93–111. allows users to create lists based on that involve them, such as pictures and stand their relationships and fine-tune geted advertising or the monetization Computing Systems (CHI 03), 2003, 10. A.-D.I. Kramer, J.E. Guillory, and J.T. features like location and workplace. posts, before these items are posted.8 their information disclosures. A learn- of social relationship information. We pp. 129–136. Hancock, “Experimental Evidence of Also, the introduction of the “close Most privacy controls apply the pri- ing mechanism would be required so must find an adequate balance between 5. “Top Sites in Social Networking,” Massive-Scale Emotional Contagion friends” and “acquaintances” lists ac- vacy preferences of the user making that these systems could understand institutional and social privacy to pro- Alexa.com; www.alexa.com/topsites through Social Networks,” Proc. Nat’l knowledges the important role of rela- the post. However, if an affected user each user’s behavior and adapt accord- tect users at both levels. /category/Computers/Internet/On Academy of Sciences, tionship strength. wants the post removed, he or she ingly to provide relevant assistance. _the_Web/Online_Communities vol. 111, no. 24, pp. 8788–8790. must negotiate with the poster using One major concern is that such systems REFERENCES /Social_Networking. 11. S. Gurses and C. Diaz, “Two Tales of RELATIONSHIP-BASED other means of communication such could become too intrusive, as they ob- 1. “Social Media Guidance for Civil 6. R. Fogues et al., “Open Challenges in Privacy in Online Social Networks,” ACCESS CONTROLS as email, texts, or private messages. serve and analyze all communications Servants: October 2014,” UK Govern- Relationship-Based Privacy Mecha- IEEE Security & Privacy, vol. 11, no. 3, Mainstream social media seems to be Even then, negotiations might happen between users. ment, 20 Oct. 2014; www.gov.uk nisms for Social Network Services,” 2013, pp. 29–37. moving toward more socially aware after the item has been already posted Except for a handful of distrib- /government/publications/social Int’l J. Human-Computer Interaction, privacy controls, which shows recog- and a privacy violation has occurred. uted social media sites that are im- -media-guidance-for-civil-servants. vol. 31, no. 5, 2015, pp. 350–370. GAURAV MISRA is a PhD student at nition of the importance of privacy in Computational mechanisms that plemented by peer-to-peer nodes, the 2. “How the Presidential Candidates 7. L. Yuan et al, “Privacy-Preserving Lancaster University. Contact him at social relationships. However, there’s automate the negotiation process provider mediates communication on Use the Web and Social Media,” Pew Photo Sharing Based on a Public Key [email protected]. still room for improvement to make have been identified as a possible mit- most of these sites. Thus, the provider Research Center: Journalism & Media, Infrastructure,” Proc. SPIE 9599, Ap- privacy controls more socially aware. igation. This doesn’t mean that us- can influence the amount and nature 15 Aug. 2012; www.journalism.org plications of Digital Processing XXX- JOSE M. SUCH is an assistant pro- Traditional access control ap- ers would lose control; instead, these of communication between individ- /2012/08/15/how-presidential VIII, 2015; doi:10.1117/12.2190458. fessor of cybersecurity in the School of Computing and Communications proaches such as role-based access mechanisms would suggest possible uals on its platforms. For example, a -candidates-use-web-and-social-media. 8. A. Lampinen et al., “We’re in It at Lancaster University. Contact him controls or group-based approaches solutions that all parties would need to recent experiment studied the emo- 3. H. Nissenbaum, “Privacy as Contex- Together: Interpersonal Manage- at [email protected]. can’t truly capture the social relation- accept. If users don’t accept the sugges- tional contagion of Facebook status tual Integrity,” Washington Law Rev., ment of Disclosure in Social Network ships among users— relationship-based tion, they need to enter into a manual updates by moderating the amount of vol. 79, 2004; www.kentlaw Services,” Proc. SIGCHI Conf. Human access controls (ReBACs) are needed.6 negotiation by other means. The main content received by users from various A ReBAC model uses a different set of challenge is to propose solutions that friends.10 This confirms that the pro- properties to define relationships, re- everyone involved will accept most of vider can modify the content being fed sulting in a more natural disclosure the time. Some mechanisms have been to the user through the network, which ACM - IEEE CS ECKERT-MAUCHLY AWARD of personal information. For example, proposed in this very young discipline, can alter the nature and strength of Keeping users can decide that only their close but a more formal study is required relationships that depend on interac- Call for Award Nominations • Deadline: 30 March 2016 • www.computer.org/awards • awards.acm.org friends can see a specific photo. to understand the conditions under tions between individuals. This also One of the main problems facing which users will make concessions, highlights the fact that institutional Stay Informed ReBACs is usability. The ReBAC model and the variables that make users take privacy—which governs the way so- YOU at the considers a large number of social stronger or more lenient positions cial media providers handle user data, Access to Computer Society books, features, which can help users better during these negotiations. including whether and how they will technical magazines and research ACM and the IEEE Computer Society co-sponsor the Eckert-Mauchly Award, whichjournals was initiated arm you in with 1979. Industry The identify the social context of their dis- Another open challenge is the abil- share data with third parties—can play award is known as the computer architecture community’s most prestigious award. intelligence to keep you ahead of the closures; however, the process of de- ity of privacy controls to adequately a role in social privacy.11 Center learning curve. fining relationships and selecting au- support users in making favorable The award recognizes outstanding contributions to computer and digital systems architecture. It comes with a certiﬁ cate and a $5,000 prize. diences needs to be easy. This entails disclosure decisions with regard to • 3,000 technical books included with membership from books 24 x improving audience visualization— self-presentation and management of ainstream social media seems The award was named for John Presper Eckert and John William Mauchly, who collaborated on the design and of Technology 7 and Safari Books Online the visual cues provided to a user relationship dynamics. Users main- to be moving toward more so- construction of the Electronic Numerical Integrator and Computer (ENIAC), the ﬁ rst large-scale electronic computing about which friends will have access tain different types of relationships cially aware privacy controls. machine, which was completed in 1947. • 13 technical magazines M • 20 research journals to the information the user posted or is with varying objectives, and the However, these efforts are still a long IEEE Computer Society about to post. Moreover, ReBACs need manner and content of their commu- way from the comprehensive modeling Questions? Write toLearn IEEE something Computer new. Society Check out TO BE PRESENTED AT to be computationally lightweight nication with these groups varies ac- of social relationships that would enable Publications Awards AdministratorComputer at [email protected] Society publications and to seamlessly integrate with the cordingly. These nuances must be ade- satisfactory boundary regulation, mim- ISCA 2016 or the ACM Awardstoday! Committee Liaison at acm social media interface, preserving the quately accommodated and supported, icking peoples’ relationships in the off- The 43rd International Symposium [email protected] dynamism of users’ social networking which poses many sociotechnical line world. Current research advances on Computer Architecture experiences. A recent effort to imple- challenges for future research on the in this area will lead to the development Nomination Guidelines: Stay relevant with• Open the to all.IEEE Anyone Computer may nominate. Society • Self-nominations are not ment a ReBAC mechanism uses at- topic. One particular challenge is the of next- generation privacy controls. accepted. • This award requires 3 endorsements. tribute-based encryption that allows development of tools that are aware These new privacy controls will, like More at www.computer.org/publications www.computer.org/computingedge 25 98 COMPUTER WWW.COMPUTER.ORG/COMPUTER MARCH 2016 99 Hua Gang Visions and Views Stevens Institute of Technology

Social-Sensed Multimedia Computing

Peng Cui and ultimedia computing, which ultimately and Malcolm Slaney organized a panel called Wenwu Zhu M aims to deliver multimedia content to “Coulda, Woulda, Shoulda: 20 Years of Multi- Tsinghua users according to their information needs media Opportunities,” in recognition of the University, China (intents), can be decomposed into various 20th anniversary of ACM Multimedia. They stages, such as multimedia compression (for invited several leading researchers in the multi- Tat-Seng Chua storage), multimedia communication (for deliv- media community to serve as panelists, includ- National ery), and multimedia content analysis (for intel- ing one of us (Ramesh Jain), along with Dick University of ligence). The first two stages are comparatively Bulterman, Larry Rowe, and Ralf Steinmetz. Singapore well studied, while multimedia analysis started Among the topics discussed, one of the most becoming mainstream in the multimedia com- thought-provoking and sobering questions was Ramesh Jain munity at the end of the last century, with why, although multimedia analysis has been a University of related technologies advancing quickly. How- hot topic in the community for dozens of years, California, Irvine ever, understanding and predicting what multi- the popular new multimedia systems and plat- media content users need in different situations forms (such as Flickr, YouTube, and Instagram) and contexts—the last-mile technology for haven’t been founded by people in the multi- delivering multimedia services—has not been media community? And why haven’t these sys- well studied (see Figure 1). tems leveraged advanced multimedia analysis This negligence has resulted in a gap between technologies? multimedia data and the information needs of Another panel at the conference, “Content is users, creating a bottleneck in advancing intelli- Dead; Long Live Content!,” further pushed gent multimedia computing technologies for attendees to be introspective about recent multi- use in real-world applications. To bridge this gap, media content-analysis research. There were var- we need to invest more effort in understanding ious, sometimes contradictory, arguments users, both individually and collectively. Fortu- during and after the conference, yet one of the nately, the explosive growth of social multime- well-accepted opinions was that researchers and dia content on the Internet is revolutionizing engineers make simple hypotheses about user the landscape of various multimedia applica- needs in terms of multimedia data. For example, tions. The new types of multimedia content, in image retrieval, a user is assumed to be look- metadata, context information, and interaction ing for images that include the query text as a behaviors in social multimedia present a signifi- tag or label; in video recommendation, a user is cant opportunity to advance and augment mul- assumed to be interested in videos that are simi- timedia and content-analysis techniques. lar to what he or she has watched. However, Existing research in this area viewed social rarely do we try to investigate and understand multimedia as a new research objective and thus the real user needs. Although relevance feedback proposed methods to exploit the new data types technology incorporates users’ interactions, the or solve the new problems.1 Here, we argue that fact that it relies on users’ explicit feedback goes the significance of social multimedia with respect against the habits of typical (lazy) users, making to the field of multimedia goes far beyond the it challenging to apply in practice. emergence of new types of data and problems. We need to treat the understanding and prediction of user needs as a first-class citizen in User-Centric Multimedia Computing the multimedia community. Now is the time to Four years ago, at the ACM International Con- reconsider the traditional multimedia comput- ference on Multimedia 2012, Klara Nahrstedt ing paradigm, which is either data centric or

26 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 92 1070-986X/16/$33.00 c 2016 IEEE Published by the IEEE Computer Society content centric. As a discipline that mainly tar- Hua Gang gets technologies and services for users, multi- Data-centric Content-centric User-centric Visions and Views Stevens Institute of Technology media computing should pay more attention to Storage Delivery Intelligent user needs. How to transform data-centric or content-centric multimedia computing into user-centric multimedia computing is a great challenge and opportunity for both academia and ? Social-Sensed Multimedia industry. Multimedia content The Semantic Gap vs. Need Gap analysis Computing Multimedia Last-mile If we say that the major goal of content-centric transmission technology multimedia computing is to bridge the seman- Peng Cui and ultimedia computing, which ultimately and Malcolm Slaney organized a panel called tic gap between low-level multimedia features Multimedia Wenwu Zhu M aims to deliver multimedia content to “Coulda, Woulda, Shoulda: 20 Years of Multi- and high-level semantics, then the ultimate compression Tsinghua users according to their information needs media Opportunities,” in recognition of the goal of user-centric computing is to bridge the University, China (intents), can be decomposed into various 20th anniversary of ACM Multimedia. They gap between multimedia content (represented Figure 1. The multimedia technology lifecycle. Understanding and predicting stages, such as multimedia compression (for invited several leading researchers in the multi- by features and semantics) and the user needs. which multimedia content users need in different situations and contexts— Tat-Seng Chua storage), multimedia communication (for deliv- media community to serve as panelists, includ- Semantics is the study of meaning. The gap the last-mile technology for delivering multimedia services—has not been well National ery), and multimedia content analysis (for intel- ing one of us (Ramesh Jain), along with Dick between the formal low-level representations in studied. University of ligence). The first two stages are comparatively Bulterman, Larry Rowe, and Ralf Steinmetz. computational machines and the richness of Singapore well studied, while multimedia analysis started Among the topics discussed, one of the most high-level semantic meanings in the human for the semantics delivered by multimedia con- becoming mainstream in the multimedia com- thought-provoking and sobering questions was mind is often referred to as the semantic gap. tents. On the other hand, the mapping from Ramesh Jain munity at the end of the last century, with why, although multimedia analysis has been a Bridging this gap—especially in unstructured semantics to user needs is complicated, and the University of related technologies advancing quickly. How- hot topic in the community for dozens of years, multimedia data, such as visual and acoustic process for learning this mapping is ill-defined. California, Irvine ever, understanding and predicting what multi- the popular new multimedia systems and plat- information—is the ultimate goal of content- Thus, between the semantics of multimedia data media content users need in different situations forms (such as Flickr, YouTube, and Instagram) centric computing technologies. Several research and the user needs for multimedia data, there and contexts—the last-mile technology for haven’t been founded by people in the multi- communities have devoted many efforts to also exists a “need” gap (see Figure 2). Both the delivering multimedia services—has not been media community? And why haven’t these sys- understanding what the data represents—for semantic gap and need gap are critical in multi- well studied (see Figure 1). tems leveraged advanced multimedia analysis example, recognizing objects in visual images, media computing, but the semantic gap has This negligence has resulted in a gap between technologies? identifying targeting events in video sequences, been the focus of more research. So can we some- multimedia data and the information needs of Another panel at the conference, “Content is and measuring textual semantic similarities at how straightforwardly derive user needs from users, creating a bottleneck in advancing intelli- Dead; Long Live Content!,” further pushed different levels (from words to documents). semantics? The answer is no. gent multimedia computing technologies for attendees to be introspective about recent multi- Although we still have a way to go in bridging It is well accepted that user needs can be rep- use in real-world applications. To bridge this gap, media content-analysis research. There were var- the gap, the success of search engine and many resented by a distribution over semantics, but we need to invest more effort in understanding ious, sometimes contradictory, arguments vertical applications, such as face detection and the distribution is heavily dependent on the users, both individually and collectively. Fortu- during and after the conference, yet one of the recognition, have demonstrated rapid advance- context. For example, the videos that a user nately, the explosive growth of social multime- well-accepted opinions was that researchers and ments toward this goal. wants to watch will depend on the user’s mood dia content on the Internet is revolutionizing engineers make simple hypotheses about user Yet once again, understanding user needs and environment, and whether the viewing is the landscape of various multimedia applica- needs in terms of multimedia data. For example, will be critical. In the scope of information sci- for work or leisure. Considering the fact that tions. The new types of multimedia content, in image retrieval, a user is assumed to be look- ence, user needs can be understood as a user’s user needs are implicit, and further considering metadata, context information, and interaction ing for images that include the query text as a desire to obtain information to satisfy his or her the incompleteness and uncertainty of observed behaviors in social multimedia present a signifi- tag or label; in video recommendation, a user is conscious or unconscious needs, and they can user behaviors, we need more comprehensive cant opportunity to advance and augment mul- assumed to be interested in videos that are simi- be further specified as interests and intents, research into discovering the mapping mecha- timedia and content-analysis techniques. lar to what he or she has watched. However, where interest represents long-term user needs nism between multimedia data and user needs Existing research in this area viewed social rarely do we try to investigate and understand and intent represents instantaneous user needs. and how this mechanism can be coupled with multimedia as a new research objective and thus the real user needs. Although relevance feedback However, these needs are often latent, so infer- rich context information. proposed methods to exploit the new data types technology incorporates users’ interactions, the ring them from observed data is challenging. A Another argument might be that if we can 2016 January–March or solve the new problems.1 Here, we argue that fact that it relies on users’ explicit feedback goes basic hypothesis is that user needs will be trig- replace multimedia data with semantics, which the significance of social multimedia with respect against the habits of typical (lazy) users, making gered in certain situations and manifested as are often textual words or sentences, then the to the field of multimedia goes far beyond the it challenging to apply in practice. behaviors. Thus, behaviors can be regarded as need-gap problem would become irrelevant. emergence of new types of data and problems. We need to treat the understanding and pre- the reflection of user needs. We argue that multimedia data cannot be fully diction of user needs as a first-class citizen in On the one hand, how users interact with abstracted by using textual semantics. Such User-Centric Multimedia Computing the multimedia community. Now is the time to multimedia (that is, their “interaction behav- semantics cannot accurately or comprehen- Four years ago, at the ACM International Con- reconsider the traditional multimedia comput- ior”) depends on the semantics of the multime- sively represent visual styles, delivered visual ference on Multimedia 2012, Klara Nahrstedt ing paradigm, which is either data centric or dia data, because users have different preferences effects, and psychovisual factors, all of which

www.computer.org/computingedge 27 92 1070-986X/16/$33.00 c 2016 IEEE Published by the IEEE Computer Society 93 Visions and Views

proactively interacting with (generating, shar- User ing, commenting on, and so forth) huge vol- umes of multimedia data. These interaction behaviors are being recorded at an unprece- Observed Behaviors dented level. Thus, social media has formed a valuable pool of data about user needs, present- Context ing a precious opportunity to bridge the need gap in multimedia computing. User needs More specifically, users’ need-related infor- Latent (Interest + Intent) mation (including long-term interests, instan- Need gap taneous intents, and emotions of both crowds and individuals), their behavior patterns, and Partially Semantics ultimately the common principles of user-mul- observed timedia interactions under different contexts Semantic gap can all be sensed from social media and summarized using social knowledge of user-multi- Observed Multimedia data media interactions. This social knowledge reflects user needs and establishes a bridge for multimedia data and user needs. Data Determining how to organically integrate multimedia data, user needs, and social knowledge into multimedia computing technology is Figure 2. The gap between the formal low-level a critical issue. Thus, we propose a new multi- representations in computational machines and media computing paradigm—social-sensed mul- the richness of high-level semantic meanings in the timedia computing—to bring social media into human mind is often referred to as the “semantic the loop of multimedia computing (see Figure gap.” Yet there is also a “need gap” between the 3). This new paradigm should naturally trans- semantics of multimedia data and the user needs form the landscape of multimedia computing for multimedia data. from traditional data-centric or content-centric multimedia computing to user-centric multi- play important roles in deciding users’ interac- media computing, which will improve users’ tion behaviors with multimedia data, especially experiences with various multimedia applica- considering the fact that most users watch mul- tions and services. timedia content for entertainment or explora- Of course, compared to traditional multime- tory goals. Therefore, both semantics and dia computing, this paradigm will face new multimedia content should be jointly consid- problems, which will be related to social multi- ered in bridging the need gap. The problem can media representation, user modeling, and user- be formulated as seeking a mapping function multimedia interaction analysis. between multimedia content (represented by both textual semantics and visual factors) and Socialized Representation of Multimedia Data the observed user behaviors. Social multimedia was first defined by Mor Naa- man as “an online source of multimedia resour- Social-Sensed Multimedia Computing ces that fosters an environment of significant Data specifying user needs is necessary to individual participation and that promotes understand real user needs for multimedia. An community curation, discussion and re-use of alternative to explicitly surveying user needs, content.”1 The main characteristic that differ- which is often costly and impractical, is to col- entiates social multimedia from traditional lect multimedia data, user data, and the interac- multimedia is that the former boasts significant tion behaviors between users and multimedia user participation and interactions with multi- data, from which we can implicitly and contin- media content. For example, users tag images uously discover users’ multimedia needs. in Flickr, add comments to videos in YouTube, Although these data types were not readily and “like” or “dislike” videos and images on available in the past, with the emergence of Facebook. These are all important resources for social media platforms (such as Flickr, Face- discovering patterns of user behaviors toward

IEEE MultiMedia book, and YouTube), billions of users are now multimedia content.

28 ComputingEdge August 2016 94 Visions and Views proactively interacting with (generating, shar- However, current multimedia representation Social ing, commenting on, and so forth) huge vol- methods (such as low-level features, concepts, User sensing umes of multimedia data. These interaction and visual attributes) are designed to bridge the Social media behaviors are being recorded at an unprece- semantic gap. There remains an obvious gap Observed Behaviors dented level. Thus, social media has formed a between semantics and user responses and valuable pool of data about user needs, present- behaviors—the need gap. Therefore, new repre- Context ing a precious opportunity to bridge the need sentation methods for social multimedia must gap in multimedia computing. be found to simplify mapping between multime- Social knowledge about user-multimedia interactions User needs More specifically, users’ need-related infor- dia content and user responses and behaviors. Latent (Interest + Intent) mation (including long-term interests, instan- Need gap taneous intents, and emotions of both crowds User Profiling and Social Graph Modeling and individuals), their behavior patterns, and User profile inference, social graph analysis, Partially Semantics ultimately the common principles of user-mul- and tie-strength measurement have become observed Multimedia User timedia interactions under different contexts popular in the social network analysis field, but Multimedia data computing needs Semantic gap can all be sensed from social media and sum- most related research has been based on text marized using social knowledge of user-multi- information. When we attempt to apply these Observed Multimedia data media interactions. This social knowledge findings and approaches to social-sensed multi- reflects user needs and establishes a bridge for media computing, several fundamental issues Figure 3. Illustration of the social-sensed multimedia computing paradigm. multimedia data and user needs. arise: Can the user and social knowledge This new paradigm should naturally transform the landscape of multimedia Data Determining how to organically integrate learned from text data be adapted to multime- computing from traditional data-centric or content-centric multimedia multimedia data, user needs, and social knowl- dia data? Is multimedia data able to tell us a dif- computing to user-centric multimedia computing. edge into multimedia computing technology is ferent (or a more complete) story about the Figure 2. The gap between the formal low-level a critical issue. Thus, we propose a new multi- characteristics of users and social relations? munication. Macroscopic analysis broadly refers representations in computational machines and media computing paradigm—social-sensed mul- Intuitively, multimedia content has intrinsi- to multimedia propagation analysis and predic- the richness of high-level semantic meanings in the timedia computing—to bring social media into cally different structures and feature spaces tion in a social environment, which can support human mind is often referred to as the “semantic the loop of multimedia computing (see Figure from text content and can typically provide multimedia popularity prediction and social gap.” Yet there is also a “need gap” between the 3). This new paradigm should naturally trans- much richer semantics and meanings. To guar- multimedia monitoring. These all make it possi- semantics of multimedia data and the user needs form the landscape of multimedia computing antee that the learned user profiles and social ble to infer user needs with regard to multimedia for multimedia data. from traditional data-centric or content-centric graph models can be seamlessly bridged with data; consequently, they play significant roles in multimedia computing to user-centric multi- multimedia data, we must revisit user profiling social-sensed multimedia computing. play important roles in deciding users’ interac- media computing, which will improve users’ and social graph modeling in the context of the tion behaviors with multimedia data, especially experiences with various multimedia applica- social multimedia environment. considering the fact that most users watch mul- tions and services. he rise of social networks and social media timedia content for entertainment or explora- Of course, compared to traditional multime- User-Multimedia Interaction Behavior Analysis T platforms brought many people to the tory goals. Therefore, both semantics and dia computing, this paradigm will face new A major goal of the sensing part of social-sensed Internet for the first time, and observable user multimedia content should be jointly consid- problems, which will be related to social multi- multimedia computing is to discover user-multi- profiles and behaviors provide us with valuable ered in bridging the need gap. The problem can media representation, user modeling, and user- media interaction behavior patterns, from resources to discover common principles of be formulated as seeking a mapping function multimedia interaction analysis. which user needs regarding multimedia data can interactions between users and multimedia between multimedia content (represented by be inferred. User-multimedia interaction behav- data. The research principles here include how both textual semantics and visual factors) and Socialized Representation of Multimedia Data iors should be investigated at different scales, and why users generate, share, and assimilate the observed user behaviors. Social multimedia was first defined by Mor Naa- depending on the level of support required by multimedia data, as outlined by Shih-Fu Chang man as “an online source of multimedia resour- various multimedia applications. In particular, and Alan Hanjalic.2,3 Notably, these common Social-Sensed Multimedia Computing ces that fosters an environment of significant these interaction behaviors can be categorized principles should not be limited to social multi- Data specifying user needs is necessary to individual participation and that promotes into the following three levels: microscopic, media platforms. Instead, they—and the understand real user needs for multimedia. An community curation, discussion and re-use of mesoscopic, or macroscopic, which correspond knowledge sensed from social media—should alternative to explicitly surveying user needs, content.”1 The main characteristic that differ- to the interaction behaviors of individual users, be generalizable to other multimedia applica-

which is often costly and impractical, is to col- entiates social multimedia from traditional groups of users, and global users, respectively. tions in which user profiles, behaviors, and 2016 January–March lect multimedia data, user data, and the interac- multimedia is that the former boasts significant Microscopic analysis can be specified as, but social relations are not observable. tion behaviors between users and multimedia user participation and interactions with multi- not limited to, user interest modeling and user For example, can knowledge sensed from data, from which we can implicitly and contin- media content. For example, users tag images sentiment analysis, which can support personal- Flickr be used to improve Google image search? uously discover users’ multimedia needs. in Flickr, add comments to videos in YouTube, ized search and recommendation for multime- Can knowledge sensed from YouTube help Although these data types were not readily and “like” or “dislike” videos and images on dia. Mesoscopic analysis includes collective design a recommendation system for TV pro- available in the past, with the emergence of Facebook. These are all important resources for behavior analysis, social influence modeling, grams? Can knowledge sensed from Flickr and social media platforms (such as Flickr, Face- discovering patterns of user behaviors toward and so on, which can support social multimedia Instagram be integrated to form a more com-

IEEE MultiMedia book, and YouTube), billions of users are now multimedia content. marketing and socially aware multimedia comprehensive understanding of users?

www.computer.org/computingedge 29 94 95 Visions and Views

The ultimate goal of social-sensed multime- 3. A. Hanjalic, “Multimedia Retrieval that Matters,” dia computing should be to sense transferable ACM Trans. Multimedia Computing, Communica- and interoperable common principles and tions, and Applications (TOMCCAP), Oct. 2013, knowledge from social multimedia and seam- article no. 44; http://dl.acm.org/citation.cfm? lessly integrate this knowledge with various doid 2523001.2490827. ¼ multimedia services. This objective poses great challenges to those who seek to improve current techniques in the multimedia community and Peng Cui is an assistant professor at Tsinghua Univer- opens up a broad assortment of new research sity, China. Contact him at [email protected]. topics that are worth investigating. MM Wenwu Zhu is a professor at Tsinghua University, China. Contact him at [email protected]. References Tat-Seng Chua is a KITHCT Chair Professor at 1. M. Naaman, “Social Multimedia: Highlighting National University of Singapore. Contact him at Opportunities for Search and Mining of Multime- [email protected]. dia Data in Social Media Applications,” Multimedia Tools and Applications, vol. 56, no. 1, 2012, pp. Ramesh Jain is a Bren Professor at the University of 9–34. California, Irvine. Contact him at [email protected]. 2. S.-F. Chang, “How Far We’ve Come: Impact of 20 Years of Multimedia Information Retrieval,” ACM Trans. Multimedia Computing, Communications, and Applications (TOMCCAP), Oct. 2013, article This article originally appeared in no. 42; http://dl.acm.org/citation.cfm? IEEE MultiMedia, vol. 23, no. 1, 2016. id 2523001.2491844. ¼ IEEE MultiMedia

30 ComputingEdge August 2016 96 Watch the World’s Leading Experts Take Multi-Core Strategies to New Heights

Purchase the IEEE Computer Society advanced Multi-Core lecture series and receive a complimentary SWEBOK Knowledge Area review course of your choice.

Please visit our website for more information about this outstanding Multi-Core Video Series. http://www.computer.org/web/education /multicore-video-series Your Homework Assignment Editor: Nargess Memarsadeghi, [email protected]

Social Resistance

Michael P. Friedlander | University of British Columbia Nathan Krislock | Northern Illinois University Ting Kei Pong | Hong Kong Polytechnic University

his case study concerns an important question in As you read through this case study, you’ll be asked to social networks: How closely related are two peo- work on a series of activities, including short mathematical ple in a network? We measure the strength of the and programming exercises, to try out these ideas on real T relationship between two people in terms of their data. You’ll need a working copy of Matlab to complete the resistance distance, a notion of distance that has its origin in programming activities. Figure 1 is a good place to start vi- the study of electrical networks.1 In that context, the resis- sualizing the core issues. tance distance measures the electrical resistance between the two nodes in a network, assuming unit resistance on Background each edge. Its relationship with the information distance In a social network, the question of how “close” two people in social networks was recently identified.2,3 (or groups) are is of great importance. The notion of close- For large networks, this distance measure can be expensive ness makes this question tricky. Take the network structure to compute. As the network changes, it’s useful to have an in- induced by Facebook friendship as an example. Such a struc- expensive way to make updates. We’ll study how various tech- ture can be conveniently modeled using an undirected graph. niques for updating matrix factorizations can help efficiently Mathematically, an undirected graph G = (V, E) consists of compute the resistance distances with small changes in the net- a set of nodes V and a set of edges E that describes connec- work. We focus on Cholesky rank-one updates, and also briefly tions between pairs of nodes. To model friendship on Face- discuss the Sherman-Morrison-Woodbury formula (as one of book, we can represent an individual by a node and join two the last exercises). We expect that you’re familiar with the idea nodes with an edge if these two people are friends on Face- of the Cholesky factorization of positive definite matrices.4 book. Figure 2 shows a small undirected graph representing We’ll guide you through a derivation of the formula for the the friendship of six individuals. Cholesky rank-one update, and introduce two aspects that may A natural measure of closeness in a graph is the length be new for you. First, you’ll have to figure out the vector used of the shortest path between two nodes. This is the least for the rank-one update when one additional connection in the number of edges that need to be traversed to get from one network is identified. Second, when matrices are large, you’ll node to another. Though well-studied mathematically, this observe that there can be a big difference in computational ef- notion isn’t an appropriate one when it comes to measuring fort when the computation involves matrix-matrix multiplica- how close two people are on a social network. For instance, tions rather than only matrix-vector multiplications. according to this measure, Jane is equally close to Alfred and Felix in Figure 2. Intuitively, however, we would expect Zip File Jane to be closer to Felix because they have two common friends. Similarly, according to the shortest path measure- ownload this companion zip file, which contains the code ment, as immediate friends, Dave is equally close to Jane Dand data you’ll use while working through the case study, and Felix. But again, we should expect Dave to be actually to the directory social-resistance: unzip(‘http://www. closer to Jane because they share two common friends. In- cs.ubc.ca/~mpf/social-resistance.zip’,‘social-resistance’); tuition tells us that our closeness measurement should obey Make sure that the directory social-resistance is in the Matlab the following rule: path, or execute the command addpath(‘social-resistance’). Two individuals are closer if they have more common friends.

32 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 98 Computing in Science & Engineering 1521-9615/16/$33.00 © 2016 IEEE Copublished by the IEEE CS and the AIP March/April 2016 Your Homework Assignment Editor: Nargess Memarsadeghi, [email protected]

0 0

500 50

1000 100 1500

150 Social Resistance 2000

2500 200 Michael P. Friedlander | University of British Columbia Nathan Krislock | Northern Illinois University 3000 Ting Kei Pong | Hong Kong Polytechnic University 250

3500 his case study concerns an important question in As you read through this case study, you’ll be asked to 300 social networks: How closely related are two peo- work on a series of activities, including short mathematical 4000 ple in a network? We measure the strength of the and programming exercises, to try out these ideas on real 0 500 1000 1500 2000 2500 3000 3500 4000 050100 150200 250300 T relationship between two people in terms of their data. You’ll need a working copy of Matlab to complete the nz = 26850 nz = 778 resistance distance, a notion of distance that has its origin in programming activities. Figure 1 is a good place to start vi- the study of electrical networks.1 In that context, the resis- sualizing the core issues. Figure 1. An author collaboration network (left) and a traffic network (right). The patterns in these figures are strikingly different: there tance distance measures the electrical resistance between seems to be relatively little structure in the collaboration network, as compared to the highly structured traffic network. the two nodes in a network, assuming unit resistance on Background each edge. Its relationship with the information distance In a social network, the question of how “close” two people in social networks was recently identified.2,3 (or groups) are is of great importance. The notion of close- For large networks, this distance measure can be expensive ness makes this question tricky. Take the network structure The resistance distance conforms with this intuitive Jane Dave to compute. As the network changes, it’s useful to have an in- induced by Facebook friendship as an example. Such a struc- definition. expensive way to make updates. We’ll study how various tech- ture can be conveniently modeled using an undirected graph. niques for updating matrix factorizations can help efficiently Mathematically, an undirected graph G = (V, E) consists of Resistance Distance Alfred Felix compute the resistance distances with small changes in the net- a set of nodes V and a set of edges E that describes connec- To describe the resistance distance, we need more no- work. We focus on Cholesky rank-one updates, and also briefly tions between pairs of nodes. To model friendship on Face- tions from graph theory. We can conveniently represent an discuss the Sherman-Morrison-Woodbury formula (as one of book, we can represent an individual by a node and join two undirected graph G = (V, E) with n nodes using an n × n Lucy Erica the last exercises). We expect that you’re familiar with the idea nodes with an edge if these two people are friends on Face- symmetric matrix A, which is called the adjacency matrix of of the Cholesky factorization of positive definite matrices.4 book. Figure 2 shows a small undirected graph representing the graph. The entries of this matrix are defined by We’ll guide you through a derivation of the formula for the the friendship of six individuals. Figure 2. A small social network of six individuals. Cholesky rank-one update, and introduce two aspects that may A natural measure of closeness in a graph is the length 1 if (,ij) ∈E Aij =  , be new for you. First, you’ll have to figure out the vector used of the shortest path between two nodes. This is the least 0 otherwise for the rank-one update when one additional connection in the number of edges that need to be traversed to get from one and the sets of nodes and edges are given by network is identified. Second, when matrices are large, you’ll node to another. Though well-studied mathematically, this where we use (i, j) to denote an edge between nodes i and j, and observe that there can be a big difference in computational ef- notion isn’t an appropriate one when it comes to measuring the set E contains a list of all edges in the graph. For example, V = {Alfred, Jane, Lucy, Dave, Erica, Felix}, fort when the computation involves matrix-matrix multiplica- how close two people are on a social network. For instance, the adjacency matrix of the graph in Figure 2 is given by (the tions rather than only matrix-vector multiplications. according to this measure, Jane is equally close to Alfred letters along the top are abbreviations for the full names) E = {(Alfred, Lucy), (Jane, Lucy), (Jane, Dave), (Jane, Erica), and Felix in Figure 2. Intuitively, however, we would expect (Lucy, Dave), (Dave, Erica), (Dave, Felix), (Erica, Felix)}. Zip File Jane to be closer to Felix because they have two common AJLDEF friends. Similarly, according to the shortest path measure- Note that this graph is undirected, that is, it doesn’t distin- 001 000 Alfred ownload this companion zip file, which contains the code ment, as immediate friends, Dave is equally close to Jane   guish between the edges (i, j) and (j, i).   Dand data you’ll use while working through the case study, and Felix. But again, we should expect Dave to be actually 0011100 Jane Let d be the vector whose ith entry is the degree of node to the directory social-resistance: unzip(‘http://www. closer to Jane because they share two common friends. In- 110100 Lucy i, that is, the number of edges attached to node i. Then A =   cs.ubc.ca/~mpf/social-resistance.zip’,‘social-resistance’);   d = Ae e d = tuition tells us that our closeness measurement should obey 011011 Dave , where is the vector of all ones. In our example, Make sure that the directory social-resistance is in the Matlab the following rule:   (1, 3, 3, 4, 3, 2). The Laplacian matrix L is defined as 010101 Erica path, or execute the command addpath(‘social-resistance’).   Two individuals are closer if they have more common friends. 000110 Felix L := Diag(d) – A,

www.computer.org/computingedge 33 98 Computing in Science & Engineering 1521-9615/16/$33.00 © 2016 IEEE Copublished by the IEEE CS and the AIP March/April 2016 www.computer.org/cise 99 Your Homework Assignment

where Diag(d) is the diagonal matrix having the vector d Activity 2 along its diagonal. For the graph in Figure 2, For the network in Figure 2, compute the resistance dis-  10−1 000   tances between Alfred and Jane, Alfred and Felix, Dave and  −−−  Jane, as well as Dave and Felix. Does this notion of distance  031110   conform with our intuition? −−1131− 00 L =   .  01−−14−−11    0101−−313 −  Our next activity computes the resistance distances in    00011−− 2 a larger network: an author collaboration network on arXiv on the subject of “General Relativity.” The data comes in the The Laplacian matrix is a symmetric matrix and is always form of an adjacency matrix: the (i, j)th entry is 1 if person i positive semidefinite—that is, xTLx ≥ 0 for all vectors x. and person j are coauthors, and zero otherwise. A graph is called connected if any two nodes in the graph are connected by a set of edges. For example, the graph in Activity 3 Figure 2 is connected; however, if the edges (Jane, Lucy) and (Dave, Lucy) are removed, then the resulting graph 1. Included in the zipped file is the matrix file Graph.mat isn’t connected because there would be no path from Alfred (a modified version of the matrix SNAP/ca-GrQc from to Felix, for example. The nullspace for the Laplacian ma- the UF Sparse Matrix Database5). This matrix is an ad- trix of a connected graph is the set of vectors that are con- jacency matrix of the collaboration network in Figure 1. stant, that is, Lv = 0 if and only if v = λe for some scalar λ. Use the following Matlab commands to load and view You will explore these properties in Activity 1, below. the matrix, as in Figure 1:

load Graph Activity 1 spy(A); 1. Prove that 2. Let L + eeT = U TU be the Cholesky factorization of L + eeT, where U is upper triangular. Show that the resis- T Le=−∑ ()()ijee() ()()ij−e() , (1) tance distances can be computed as (,ij)∈E 2 (i) −Ti()() ()j where e is the vector that is one in the ith entry, and is rUij =−ee, zero in every other entry. 2. Use the relationship in Equation 1 to show that the where U–T = (U–1)T = (U T)–1. Based on this, write a Matlab Laplacian matrix is positive semidefinite. function to compute rij for any given i and j. To compute z = 3. Show that Le = 0. U–T(e(i) – e(j)) efficiently, you should solve thelower-triangular 4. Let G be a connected graph and let L be its Laplacian system U Tz = b, with b = e(i) – e( j), using forward substitution; T matrix. Show that L + ee is positive definite, that is, that this is done in Matlab using the command z = U’\b. Don’t xT(L + eeT )x > 0 for all nonzero vectors x. be tempted to use the Matlab command z = inv(U’)*b! This is less efficient and less accurate.

[rij,U] = Resistance(Lones,i,j); The resistance distance between nodes i and j on a connected graph is defined as The inputs are the positive definite matrix Lones (corresponding to L + eeT in the problem), i and j. The first output ()ij()2 reij :,=−e M argument rij is the resistance distance rij, and the second output argument U is the Cholesky factor of Lones. You can use T –1 where M = (L + ee ) , and ||x||M is the weighted 2-norm of the Matlab built-in function chol to compute U. Compare a vector x, defined by your output against the solution fileResistance_soln.p. 3. Complete the missing sections of code in the file T xxM = Mx . changeResistance.m:

r = changeResistance(A,i,j,k); . As required for || ||M to be a valid vector norm, M is T positive definite because L + ee is positive definite. Note This code takes in the adjacency matrix A, i, and j, that the usual 2-norm of a vector corresponds to M = I, that and also a positive integer k. It connects i to at most k T is, xx≡=I xx. random neighbors of j one by one, and recomputes the

34 ComputingEdge August 2016 100 March/April 2016 Your Homework Assignment

where Diag(d) is the diagonal matrix having the vector d Activity 2 resulting resistance distance between i and j after each Then let Q1 be the Givens matrix along its diagonal. For the graph in Figure 2, addition. The k computed resistances are recorded in the cs0 −  For the network in Figure 2, compute the resistance dis- output vector r.    10−1 000 i = j = Q =  I    tances between Alfred and Jane, Alfred and Felix, Dave and Experiment with the values 500, 3000, and 1100n−  ,  −−−  Jane, as well as Dave and Felix. Does this notion of distance k = 10. Compare your output against the solution change-    031110 sc0    conform with our intuition? Resistance_soln.p. You can use rng(‘default’) to re- −−1131− 00 L =   . set the random seed for comparing the two codes. where In–1 is the (n – 1) × (n – 1) identity matrix, and 0 is  01−−14−−11   Plot the decrease in the resistance distance as k used to represent a vector or matrix of zeros with the appro-  0101−−313 −  Our next activity computes the resistance distances in increases. This indicates, intuitively, if researcher i works priate dimension. Multiply the matrix in Equation 2 by Q ,   1  00011−− 2 a larger network: an author collaboration network on arXiv with coauthors of researcher j, then i and j get “closer” to and “zero out” the first entry of the last row: on the subject of “General Relativity.” The data comes in the each other on this collaboration network.    2 2  The Laplacian matrix is a symmetric matrix and is always form of an adjacency matrix: the (i, j)th entry is 1 if person i U11 *  U11 + xx1 * ()1 U     U  positive semidefinite—that is, xTLx ≥ 0 for all vectors x. and person j are coauthors, and zero otherwise.         Q11= Q 0 * =  0 * =: .  T     ()1 T  connected x    ()x  A graph is called if any two nodes in the graph  x *  0 *   are connected by a set of edges. For example, the graph in Rank-One Updates  1    Activity 3 Figure 2 is connected; however, if the edges (Jane, Lucy) In the above activity, we performed a Cholesky factorization The symbol * is used as a wild card to denote a pos- and (Dave, Lucy) are removed, then the resulting graph 1. Included in the zipped file is the matrix file Graph.mat each time a collaboration is added. The overall work can be sibly nonzero entry. Because of the position of In–1 in Q1, isn’t connected because there would be no path from Alfred (a modified version of the matrix SNAP/ca-GrQc from significantly reduced by using a Cholesky rank-one update, it only affects the first and last rows of the matrix that it to Felix, for example. The nullspace for the Laplacian ma- the UF Sparse Matrix Database5). This matrix is an ad- as we describe next. multiplies. trix of a connected graph is the set of vectors that are con- jacency matrix of the collaboration network in Figure 1. Suppose that B is an n × n positive definite matrix and stant, that is, Lv = 0 if and only if v = λe for some scalar λ. Use the following Matlab commands to load and view B = UTU is its Cholesky factorization. Given this upper tri- Activity 5 You will explore these properties in Activity 1, below. the matrix, as in Figure 1: angular matrix U and a vector x, we would like to obtain a Cholesky factorization for B + xxT. Note that 1. Verify that Q is orthogonal, that is, show that QQT is load Graph 1 11 +  U  the identity matrix of size (n 1). Then show that Activity 1 spy(A); TT TT[]  Bx+=xUUx +=xUx T , T T x  + =   ()1 T ()1 1. Prove that 2. Let L ee U U be the Cholesky factorization of  U   U  L + eeT, where U is upper triangular. Show that the resis- where the matrix     ==+T T  T   T  Bxx . ()()ij() ()()ij() ()()1 ()()1 Le=−∑ ee−e , (1) tance distances can be computed as  U   x   x  (,ij)∈E   (2) 2  T  (i) −Ti()() ()j x  where e is the vector that is one in the ith entry, and is rUij =−ee, A matrix Q2 is then similarly chosen to operate on the zero in every other entry. is of size (n + 1) × n. second and last rows of –T –1 T T –1 2. Use the relationship in Equation 1 to show that the where U = (U ) = (U ) . Based on this, write a Matlab  U ()1  =   Laplacian matrix is positive semidefinite. function to compute rij for any given i and j. To compute z  T  –T (i) (j) Activity 4 ()()1 3. Show that Le = 0. U (e – e ) efficiently, you should solve thelower-triangular  x  4. Let G be a connected graph and let L be its Laplacian system U Tz = b, with b = e(i) – e( j), using forward substitution; Let a and b be given scalars, with a > 0. Find (c, s) with c > so as to zero out the second entry of the last row, and so T matrix. Show that L + ee is positive definite, that is, that this is done in Matlab using the command z = U’\b. Don’t 0 and c2 + s2 = 1 so that on. Thus, after k iterations we have xT(L + eeT )x > 0 for all nonzero vectors x. be tempted to use the Matlab command z = inv(U’)*b! This cs− a  22    ab+  ()k is less efficient and less accurate.    =  . (3)  U  U  scb          0   T  = QQk ...21Q T . [rij,U] = Resistance(Lones,i,j); ()()k x  The resistance distance between nodes i and j on a con-  x    nected graph is defined as The inputs are the positive definite matrix Lones (corresponding to L + eeT in the problem), i and j. The first output The 2-by-2 matrix in Equation 3 is an example of a Complete the missing sections of code in the file ()ij()2 reij :,=−e M argument rij is the resistance distance rij, and the second out- Givens rotation. One way to perform a Cholesky rank-one mycholupdate.m. This function takes the above approach put argument U is the Cholesky factor of Lones. You can use update is to apply successive Givens rotation matrices Q to to compute the Cholesky factor of B + xxT, given U and x. where M = (L + eeT )–1, and ||x|| is the weighted 2-norm of the Matlab built-in function chol to compute U. Compare “zero out” the last row of the matrix in Equation 2 step by M U = mycholupdate(U,x); a vector x, defined by your output against the solution fileResistance_soln.p. step and obtain a new U in the process: the Cholesky factor 3. Complete the missing sections of code in the file of B + xxT. Remember, you don’t need to form the whole Givens T xxM = Mx . changeResistance.m: Using the formulas you discover in Activity 4, you can matrix; in each iteration, you only need to find c( , s) with find (c, s) with c > 0 and c2 + s2 = 1 such that c > 0 and c2 + s2 = 1 such that r = changeResistance(A,i,j,k); . As required for || ||M to be a valid vector norm, M is T ()i−1  2 2  positive definite because L + ee is positive definite. Note This code takes in the adjacency matrix A, i, and j, cs− U   2 2  cs− U   ()i−−1 ()i 1   11  Ux11 + 1    ii   ()U iii +()xi  that the usual 2-norm of a vector corresponds to M = I, that and also a positive integer k. It connects i to at most k    =  .  −  =  .      sc ()i 1    T sc x1   0    xi   0  is, xx≡=I xx. random neighbors of j one by one, and recomputes the

www.computer.org/computingedge 35 100 March/April 2016 www.computer.org/cise 101 Your Homework Assignment

Test your code with random instances: In this code, we assume knowledge of the Cholesky factor U of L + eeT, and compare the additional cost needed to B = randn(3000); B = B’*B; obtain y from performing the Cholesky update or using the U = chol(B); x = randn(3000,1); SMW formula. The outputserrchol and errsmw are residuals Compare your code against the solution mycholupdate_ soln.p, and then compare your solution to Matlab’s built-in ()Le++exTTxy−−()ee()ij() , function cholupdate.m. with the y obtained from the corresponding approach. 2. Complete the missing sections of code in the file Compare your code against the solution filechol_vs_ changeResistance_r1.m: smw_soln.p. You can also use Matlab’s built-in function cholupdate.m in place of mycholupdate.m. r = changeResistance_r1(A,i,j,k);

It has the same inputs and outputs as change Resistance.m, and it uses the Cholesky rank-one update The SMW formula is convenient, but for some matri- in the loop instead of computing Cholesky factors afresh. ces, it can be unstable—that is, roundoff errors intrinsic to Guided by Activity 1.1, you’ll need to figure out what the a computer’s finite-precision arithmetic can overwhelm the x is when an edge is added to the graph. final calculated answer.6 It doesn’t happen often, but it’s Experiment with your code using i = 500, j = 3000, something to bear in mind. and k = 10. Compare your output against the solution In our last activity, we’re going to apply what we learned filechangeResistance_r1_soln.p. How does it compare above to analyze the traffic network shown in the panel on with the method in Activity 3.3 in terms of speed? You can the right-hand side of Figure 1. also try to use the Matlab built-in function cholupdate.m in place of mycholupdate.m. Activity 7 1. Inside the zip file is the matrix file Graph2.mat, which is a T he Sherman-Morrison-Woodbury Formula modified version of the matrix SNAP/roadNet-CA from the Another widely used method for obtaining (B + xxT)–1 UF Sparse Matrix Database.5 This is the adjacency matrix of given B–1 is the Sherman-Morrison-Woodbury (SMW) for- the traffic network in Figure 1 and represents a subset of the mula. The SMW formula states that road network of California. Load and view the matrix as in Activity 3.1. −1 − 1 −− ()Bx+ xBT =−1 Bx11xBT . Compute the resistance distances for all pairs of nodes. 1+ xBT −1x What’s the average resistance distance across the whole network? Plot a histogram of the resistance distances. The formula is particularly useful in our scenario because we only need to find the matrix-vector product B( + xxT )–1v 2. As a greedy approach to decrease the average resistance for a suitable B, x, and v. distance, you can build a new road between the pair of nodes i and j that have the largest resistance distance. What’s the average resistance distance across the Activity 6 whole network after such a new road is built? You can use 1. Argue that if B–1 is given, then computing (B + xxT)–1v either the Cholesky rank-one update or the SMW formula for given vectors x and v using the SMW formula requires to compute the new resistance distances. You might also at most Cn2 additions and multiplications, for some con- want to test the effectiveness of the greedy approach by stant C independent of n. comparing the resulting average resistance distance with that obtained by adding a new road between two randomly 2. Complete the missing sections of code in the file chol_vs_ chosen nodes. smw.m. This function computes

−1 yL=+()eeTT+ xx ()ee()ij− () , e’ve only touched on a few of the basic methods in where x corresponds to adding an edge between node i W numerical linear algebra needed to analyze networks. and a random neighbor of j: With just a few refinements, it’s possible to apply the techniques that you’ve learned to much larger networks. Can [errchol,errsmw] = chol_vs_smw(A,i,j); you spot opportunities to make your code more efficient?

36 ComputingEdge August 2016 102 March/April 2016 Your Homework Assignment

Test your code with random instances: In this code, we assume knowledge of the Cholesky fac- For example, in Activities 3 and 6, you computed the of SIAM J. Optimization, SIAM J. Matrix Analysis and Appli- tor U of L + eeT, and compare the additional cost needed to Cholesky factorization of L + eeT by first explicitly form- cations, Mathematics of Operations Research, and Mathematical B = randn(3000); B = B’*B; obtain y from performing the Cholesky update or using the ing this matrix. But the very same techniques you studied Programming. He received a PhD in operations research from U = chol(B); x = randn(3000,1); SMW formula. The outputserrchol and errsmw are residuals for updating the Cholesky factorization could be used: Stanford University. Contact him at [email protected]. Compare your code against the solution mycholupdate_ first, compute the Cholesky factorization UTU = L, and soln.p, and then compare your solution to Matlab’s built-in ()Le++exTTxy−−()ee()ij() , then obtain the required factorization of L + eeT via U = Nathan Krislock is an assistant professor at Northern Illinois function cholupdate.m. mycholupdate(U,e). University. His research interests include continuous and combi- with the y obtained from the corresponding approach. We also haven’t touched on a crucial property of most natorial optimization, with a focus on semidefinite programming, 2. Complete the missing sections of code in the file Compare your code against the solution filechol_vs_ networks that arise in practice: they have relatively few edg- numerical computation, and applications. Krislock received a PhD changeResistance_r1.m: smw_soln.p. You can also use Matlab’s built-in function es as compared to nodes, which means that the adjacency in mathematics from the University of Waterloo. Contact him at cholupdate.m in place of mycholupdate.m. and Laplacian matrices that we studied are sparse (that is, [email protected]. r = changeResistance_r1(A,i,j,k); they have very few nonzero entries). The traffic network It has the same inputs and outputs as change from which Figure 1 is extracted exemplifies this property. Ting Kei Pong is an assistant professor at Hong Kong Polytech- Resistance.m, and it uses the Cholesky rank-one update The SMW formula is convenient, but for some matri- The original network, which has 19,171,281 nodes, is too nic University. His research is mainly on continuous optimization, in the loop instead of computing Cholesky factors afresh. ces, it can be unstable—that is, roundoff errors intrinsic to large to treat with the dense-matrix techniques used in this with a focus on first-order methods for large-scale problems. Pong Guided by Activity 1.1, you’ll need to figure out what the a computer’s finite-precision arithmetic can overwhelm the case study but could be easily stored and manipulated by received a PhD in mathematics from the University of Washing- x is when an edge is added to the graph. final calculated answer.6 It doesn’t happen often, but it’s taking advantage of sparsity. ton. Contact him at [email protected]. Experiment with your code using i = 500, j = 3000, something to bear in mind. Numerical linear algebra is a cornerstone of countless and k = 10. Compare your output against the solution In our last activity, we’re going to apply what we learned practical problems, both old and new. We hope this case filechangeResistance_r1_soln.p. How does it compare above to analyze the traffic network shown in the panel on study has piqued your interest! with the method in Activity 3.3 in terms of speed? You can the right-hand side of Figure 1. ThisSelected article articles originally and appearedcolumns from in IEEE Computer Society also try to use the Matlab built-in function cholupdate.m Acknowledgments Computingpublications in areScience also available & Engineering for free ,at vol. http://Computing 18, - in place of mycholupdate.m. We extend sincere thanks to Dianne O’Leary and Nargess Me- Now.computer.org.no. 2, 2016. Activity 7 marsadeghi for their steady encouragement and their many 1. Inside the zip file is the matrix file Graph2.mat, which is a thoughtful suggestions for improving the presentation. This work The Sherman-Morrison-Woodbury Formula modified version of the matrix SNAP/roadNet-CA from the was funded in part by a grant from the Carl Wieman Science Edu- marcH/aprIL 2013 www.computer.org/software Another widely used method for obtaining (B + xxT)–1 UF Sparse Matrix Database.5 This is the adjacency matrix of cation Initiative at the University of British Columbia. may/June 2013 www.computer.org/software given B–1 is the Sherman-Morrison-Woodbury (SMW) for- the traffic network in Figure 1 and represents a subset of the mula. The SMW formula states that road network of California. Load and view the matrix as in References Activity 3.1. 1. D. Klein and M. Randic, “Resistance Distance,” J. Math- − ()T 1 −1 1 −−11T Compute the resistance distances for all pairs of nodes. ematical Chemistry, vol. 12, 1993, pp. 81–95. Bx+ xB=− − BxxB . 1+ xBT 1x What’s the average resistance distance across the whole 2. E. Bozzo and M. Franceschet, “Resistance Distance, Close- network? Plot a histogram of the resistance distances. ness, and Betweenness,” Social Networks, vol. 35, no. 3, 2013, The formula is particularly useful in our scenario because pp. 460–469. we only need to find the matrix-vector product B( + xxT )–1v 2. As a greedy approach to decrease the average resistance 3. K. Stephenson and M. Zelen, “Rethinking Centrality: Meth- gotfrom minecraft to minds // 11 flaws? B x v ods and Examples,” Social Networks, vol. 11, no. 1, 1989, for a suitable , , and . distance, you can build a new road between the pair of Landing a spacecraft on mars // 83

Design patterns: magic or myth? // 87 nodes i and j that have the largest resistance distance. pp. 1–37. storytelling for software professionals // 9 What’s the average resistance distance across the 4. L.N. Trefethen and D. Bau III, Numerical Linear Algebra, In Defense of Boring // 16 Activity 6 Beyond Data mining // 92 whole network after such a new road is built? You can use SIAM, 1997. 1. Argue that if B–1 is given, then computing (B + xxT)–1v either the Cholesky rank-one update or the SMW formula 5. T.A. Davis and Y. Hu, “The University of Florida Sparse for given vectors x and v using the SMW formula requires to compute the new resistance distances. You might also Matrix Collection,” ACM Trans. Mathematical Software, vol. at most Cn2 additions and multiplications, for some con- want to test the effectiveness of the greedy approach by 38, no. 1, 2011, pp. 1:1–1:25; www.cise.ufl.edu/research/ stant C independent of n. comparing the resulting average resistance distance with sparse/matrices. that obtained by adding a new road between two randomly 6. E.L. Yip, “A Note on the Stability of Solving a Rank-p Modi- IEEE Software offers pioneering ideas, expert 2. Complete the missing sections of code in the file chol_vs_ chosen nodes. fication of a Linear System by the Sherman-Morrison-Wood- smw.m. This function computes bury Formula,” SIAM J. Scientific and Statistical Computing, analyses,Find and out thoughtful more and insightsget involved: for software vol. 7, no. 2, 1986, pp. 507–513. professionalscybersecurity.ieee.org who need to keep up with −1 yL=+()eeTT+ xx ()ee()ij− () , rapid technology change. It’s the authority on e’ve only touched on a few of the basic methods in translating software theory into practice. where x corresponds to adding an edge between node i W numerical linear algebra needed to analyze networks. Michael P. Friedlander is a professor of computer science and and a random neighbor of j: With just a few refinements, it’s possible to apply the tech- mathematics at the University of British Columbia. His research is www.computer.org/software/subscribe niques that you’ve learned to much larger networks. Can primarily in developing and implementing numerical methods for [errchol,errsmw] = chol_vs_smw(A,i,j); you spot opportunities to make your code more efficient? large-scale optimization. Friedlander serves on the editorial boards

www.computer.org/computingedge 37 102 March/April 2016 www.computer.org/cise 103 Securing iT eDiTOrS: rick Kuhn, uS national institute of Standards and Technology, [email protected] Tim Weil, Scram Systems, [email protected]

Spammers Are Becoming “Smarter” on Twitter

Chao Chen, Jun Zhang, Yang Xiang, and Wanlei Zhou, Deakin University, Australia Jonathan Oliver, Trend Micro

witter has become one of combat spam, the percentage of use @ to make spam tweets ap- the most commonly used spam in the whole platform is still pear on the victim’s feed without communication tools in high. We hypothesize that this is being a follower of the victim—for daily life. With 500 mil- because spammers are becoming example, a spam tweet will appear Tlion users, Twitter now generates more cunning on Twitter. While on Barack Obama’s timeline if it is more than 500 million tweets per researchers are developing meth- written with @obama. By embed- day. However, its popularity has ods to detect spam, spammers ding popular hashtag keywords, also attracted spamming. Spam- continuously invent new strategies one spam tweet can become part mers spread many intensive tweets, to bypass detection. of a trending topic that can then which can lure legitimate users to Here, we review the well-known be viewed by a victim who is inter- commercial or malicious sites con- methods that spammers have used ested in that topic. For example, taining malware downloads, phish- to avoid or reduce their chances of a spam tweet with #007 will be ing, drug sales, scams, and more.1 being caught on Twitter. We show disseminated to victims who are Spam is a problem through- that spammers are now using browsing the popular James Bond out the Internet, and Twitter is more advanced strategies, namely book and film series. Spammers not immune. In addition, Twit- coordinated posting behavior, finite- also use other Twitter functions, ter spam is much more successful state-machine-based spam template, such as “reply,” “favorite,” and compared to email spam.2 Various and passive spam. “following” to spread spam.7 For- methods have been proposed by tunately, researchers can also use researchers to deal with Twitter Well-Known Spamming these features (such as the num- spam, such as identifying spam- Strategies ber of followers or the number of mers based on tweeting history3 At the most basic level, spam- hashtags) to detect Twitter spam.3 or social attributes,4 detecting mers use various Twitter func- To bypass such detection sys- abnormal behavior,5 and classify- tions such as @ and hashtags (#) tems, spammers apply evasion ing tweet-embedded URLs.6 Al- to engage victims (see the “Twitter tactics, such as gaining more fol- though researchers, as well as Features” sidebar for a breakdown lowers, posting more tweets, and Twitter itself, have attempted to of Twitter terms). Spammers can so on.8 They aren’t exposed by the

38 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 66 IT Pro March/April 2016 Published by the IEEE Computer Society 1520-9202/16/$33.00 © 2016 IEEE Securing iT eDiTOrS: rick Kuhn, uS national institute of Standards and Technology, [email protected] Twitter Features Tim Weil, Scram Systems, [email protected] simple detection systems already The following Twitter features were relevant to our work: described because their activity ➤ Mention (@): If a tweet contains the @user tag, it is called a men- mimics that of legitimate users. To tion. The mentioned tweet can appear in a user’s timeline even if combat this, researchers propose the sender is neither followed by nor follows the user. robust social graph-based features, ➤ Hashtag (#): A hashtag embedded in a tweet is normally a keyword such as local clustering coefficients, Spammers 8 to describe this tweet. If the hashtagged keyword is very popular, it betweenness centrality, and dis- will become a trending topic that can be seen by all Twitter users. tance/connectivity4 to detect those ➤ Timeline: The tweets sent by those a user follows or tweets that use tweets fabricated by spammers. Are Becoming @ to mention the user appear in the user’s timeline. In addition to the aforemen- tioned spamming strategies, our “Smarter” on Twitter spam analysis reveals that spammers are now using more advanced methods. Table 1. Spam breakdown. Twitter Coordinated Posting Group Spam type Spam tweets (%) Behavior We collected a dataset of more A Edu and so on 27. 28 than 570 million tweets with B Cracked software, games 8.11 URLs from 25 September 2013 to C Edu 6.26 Chao Chen, Jun Zhang, Yang Xiang, and Wanlei Zhou, 9 October 2013. Within this data- D Cracked 6.19 Deakin University, Australia set, we identified around 33 million E Cracked software 4.39 Jonathan Oliver, Trend Micro spam tweets using Trend Micro’s F Printer, mobile 3.72 Web Reputation Technology;9 this G Twitter follower 2.54 accounts for 5.8 percent of the total tweets. We then used bipartite H Video, mobile, cracked software, games 2.23 cliques to cluster the spam tweets I Games, computer 2.04 witter has become one of combat spam, the percentage of use @ to make spam tweets ap- into 17 groups, as shown in Table 1 J Edu and so on 1.99 the most commonly used spam in the whole platform is still pear on the victim’s feed without (see the “Bipartite Cliques” side- K Shirt 1.91 communication tools in high. We hypothesize that this is being a follower of the victim—for bar for details about this meth- L Games, mobile printer 1.81 daily life. With 500 mil- because spammers are becoming example, a spam tweet will appear od). Seventeen groups dominate M Computer, printer 1.77 Tlion users, Twitter now generates more cunning on Twitter. While on Barack Obama’s timeline if it is more than 75 percent of the spam, more than 500 million tweets per researchers are developing meth- written with @obama. By embed- whereas “others” account for less N Games, hardware spam 1.53 day. However, its popularity has ods to detect spam, spammers ding popular hashtag keywords, than 25 percent, indicating that, in O Computer game, mobile device 1.41 also attracted spamming. Spam- continuously invent new strategies one spam tweet can become part general, spam is sent by groups. P Credit-card, edu 1.08 mers spread many intensive tweets, to bypass detection. of a trending topic that can then We also found that six groups in Q Cracked software, games 1.02 which can lure legitimate users to Here, we review the well-known be viewed by a victim who is inter- Table 1 (groups A, B, C, E, I, and Other spam 24.74 commercial or malicious sites con- methods that spammers have used ested in that topic. For example, J, the bold, italicized letters in the We named spam types according to the spam’s content—for example, cracked software taining malware downloads, phish- to avoid or reduce their chances of a spam tweet with #007 will be table) had some common features: spam is about cracked software. ing, drug sales, scams, and more.1 being caught on Twitter. We show disseminated to victims who are Spam is a problem through- that spammers are now using browsing the popular James Bond • The URLs embedded in the out the Internet, and Twitter is more advanced strategies, namely book and film series. Spammers tweets tend to use a .ru (server not immune. In addition, Twit- coordinated posting behavior, finite- also use other Twitter functions, of Russian origin) domain. ter spam is much more successful state-machine-based spam template, such as “reply,” “favorite,” and • The content of the landing pag- compared to email spam.2 Various and passive spam. “following” to spread spam.7 For- es were written in Russian. Bipartite Clique methods have been proposed by tunately, researchers can also use • The URLs tended to end with a researchers to deal with Twitter Well-Known Spamming these features (such as the num- Unix time stamp, such as http:// o identify the bipartite cliques, we first extracted the domains of spam, such as identifying spam- Strategies ber of followers or the number of xxxx.ru/xxxx-1380642617.html. TURLs embedded in tweets along with those tweets’ senders. Then, 3 3 mers based on tweeting history At the most basic level, spam- hashtags) to detect Twitter spam. we constructed a graph in which the Twitter users were nodes on one 4 or social attributes, detecting mers use various Twitter func- To bypass such detection sys- To study the spamming behavior side of the graph while the domains in sent tweets were nodes on the 5 abnormal behavior, and classify- tions such as @ and hashtags (#) tems, spammers apply evasion of these six groups, we counted the other side. For each tweet from user U that contained a link with do- 6 ing tweet-embedded URLs. Al- to engage victims (see the “Twitter tactics, such as gaining more fol- tweets sent per hour by each group. main D, we connected this user U to domain D in the graph. Once the though researchers, as well as Features” sidebar for a breakdown lowers, posting more tweets, and Group A spread spam actively from graph was fully connected, a bipartite clique was formed. Twitter itself, have attempted to of Twitter terms). Spammers can so on.8 They aren’t exposed by the 26 September (see Figure 1). When

140,000 120,000 Group A Group B Group C Group E Group J Group I 100,000 80,000 60,000 40,000 20,000 0 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 2013-10-08: 11 2013-10-07: 11 2013-10-06: 11 2013-10-05: 11 2013-10-04: 11 2013-10-03: 11 2013-10-02: 11 2013-10-01: 11 2013-09-30: 11 2013-09-29: 11 2013-09-28: 11 2013-09-27: 11 2013-09-26: 11 2013-09-25: 11 2013-10-08:15:00 2013-10-08:19:00 2013-10-08:23:00 2013-10-09:03:00 2013-10-09:07:00 2013-10-07:15:00 2013-10-07:19:00 2013-10-07:23:00 2013-10-08:03:00 2013-10-08:07:00 2013-10-06:15:00 2013-10-06:19:00 2013-10-06:23:00 2013-10-07:03:00 2013-10-07:07:00 2013-10-05:15:00 2013-10-05:19:00 2013-10-05:23:00 2013-10-06:03:00 2013-10-06:07:00 2013-10-04:15:00 2013-10-04:19:00 2013-10-04:23:00 2013-10-05:03:00 2013-10-05:07:00 2013-10-03:15:00 2013-10-03:19:00 2013-10-03:23:00 2013-10-04:03:00 2013-10-04:07:00 2013-10-02:15:00 2013-10-02:19:00 2013-10-02:23:00 2013-10-03:03:00 2013-10-03:07:00 2013-10-01:15:00 2013-10-01:19:00 2013-10-01:23:00 2013-10-02:03:00 2013-10-02:07:00 2013-09-30:15:00 2013-09-30:19:00 2013-09-30:23:00 2013-10-01:03:00 2013-10-01:07:00 2013-09-29:15:00 2013-09-29:19:00 2013-09-29:23:00 2013-09-30:03:00 2013-09-30:07:00 2013-09-28:15:00 2013-09-28:19:00 2013-09-28:23:00 2013-09-29:03:00 2013-09-29:07:00 2013-09-27:15:00 2013-09-27:19:00 2013-09-27:23:00 2013-09-28:03:00 2013-09-28:07:00 2013-09-26:15:00 2013-09-26:19:00 2013-09-26:23:00 2013-09-27:03:00 2013-09-27:07:00 2013-09-25:15:00 2013-09-25:19:00 2013-09-25:23:00 2013-09-26:03:00 2013-09-26:07:00 2013-09-25:07:00

Figure 1. The number of spam tweets sent by the six groups highlighted in Table 1. (Note that data were lost for 29 and 30 September).

nuts (240)

rofl (270) odd (240) omfg (270) funny (240) tweet (432) lol (270) epic (240) post (432) you (720) is (1,080) lmao (270) this (2,160) update (432) by (2,160) u (720) 7 awesome (240) 2 3 4 5 6 was (1,080) blog (432) yours (720) crazy (240) 8 1 lmfao (270) SHORT_URL (2,160) USER (2,160) LOL (270) entry (432) cool (240) lolz (270) hilarious (240) haha (270) 0 so funny (240) 9 $ (2,160)

Figure 2. Finite-state machine-based spam template. Spammers can use this template to generate 2,160 different spam tweets.

group A stopped sending spam, “celebrity name” + “an eye-catching different tweets. Take the finite- group C started sending it on 4 action” + URL. Therefore, research- state machine in Figure 2, for ex- October. Groups C and E, and E ers can extract the templates and ample; it has 8 × 5 × 3 × 2 × 9 = and J also displayed this type of match tweets to them to detect spam. 2,160 different routes from start to spamming behavior. We regard We found that spammers are end. This means that spammers this behavior as coordinated post- now using more complex tem- can use this template to generate ing behavior, a phenomenon in plates to generate spam. Sur- 2,160 different spam tweets with which one group of spam tweets prisingly, spammers are using little effort. For example, spam- disappears while another is being finite-state machines to generate mers can write a script that ran- sent. This kind of posting behavior what we have named finite-state domly chooses one option from is more difficult to detect because machine-based spam templates each node to generate one spam spammers change the groups of (see Figure 2). One finite-state tweet. Relying on simple string accounts to abuse Twitter. machine has a number of states, signatures to match spam tweets and each edge of it is denoted by will allow most of these finite- Finite-State Machine- one word. If we travel from the state machine-based template Based Spam Template beginning to the end, we can spam tweets to escape detection. Some have found that most spam is have one full sentence, such as generated using specific templates,10 “lol, this tweet by you is funny + Passive Spam which is logical because it is very ex- SHORT URL” in the finite-state As previously described, traditional pensive for spammers to write each machine. By using one finite-state spam is distributed using Twitter tweet manually. However, the tem- machine-based spam template, functions such @ and #. However, plate is often simple10—for example, spammers can generate many we also found that much spam

40 ComputingEdge August 2016 68 IT Pro March/April 2016 securing it

does not use any tags. As a result, 140,000 such spam cannot be identified 120,000 Group A Group B Group C Group E Group J Group I by machine-learning-based spam 100,000 detection that uses these features. 80,000 Contrary to traditional spam, 60,000 which tries to involve victims as 40,000 much as possible, this spam is only 20,000 0 viewed by victims when they search :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 :00 for specific key words. Consequent- ly, we call this passive spam. None 2013-10-08: 11 2013-10-07: 11 2013-10-06: 11 2013-10-05: 11 2013-10-04: 11 2013-10-03: 11 2013-10-02: 11 2013-10-01: 11 2013-09-30: 11 2013-09-29: 11 2013-09-28: 11 2013-09-27: 11 2013-09-26: 11 2013-09-25: 11 2013-10-08:15:00 2013-10-08:19:00 2013-10-08:23:00 2013-10-09:03:00 2013-10-09:07:00 2013-10-07:15:00 2013-10-07:19:00 2013-10-07:23:00 2013-10-08:03:00 2013-10-08:07:00 2013-10-06:15:00 2013-10-06:19:00 2013-10-06:23:00 2013-10-07:03:00 2013-10-07:07:00 2013-10-05:15:00 2013-10-05:19:00 2013-10-05:23:00 2013-10-06:03:00 2013-10-06:07:00 2013-10-04:15:00 2013-10-04:19:00 2013-10-04:23:00 2013-10-05:03:00 2013-10-05:07:00 2013-10-03:15:00 2013-10-03:19:00 2013-10-03:23:00 2013-10-04:03:00 2013-10-04:07:00 2013-10-02:15:00 2013-10-02:19:00 2013-10-02:23:00 2013-10-03:03:00 2013-10-03:07:00 2013-10-01:15:00 2013-10-01:19:00 2013-10-01:23:00 2013-10-02:03:00 2013-10-02:07:00 2013-09-30:15:00 2013-09-30:19:00 2013-09-30:23:00 2013-10-01:03:00 2013-10-01:07:00 2013-09-29:15:00 2013-09-29:19:00 2013-09-29:23:00 2013-09-30:03:00 2013-09-30:07:00 2013-09-28:15:00 2013-09-28:19:00 2013-09-28:23:00 2013-09-29:03:00 2013-09-29:07:00 2013-09-27:15:00 2013-09-27:19:00 2013-09-27:23:00 2013-09-28:03:00 2013-09-28:07:00 2013-09-26:15:00 2013-09-26:19:00 2013-09-26:23:00 2013-09-27:03:00 2013-09-27:07:00 2013-09-25:15:00 2013-09-25:19:00 2013-09-25:23:00 2013-09-26:03:00 2013-09-26:07:00 2013-09-25:07:00 of these spam tweets have tags embedded (see Figure 3), and they are Figure 1. The number of spam tweets sent by the six groups highlighted in Table 1. (Note that data were lost mostly promoting cracked games, for 29 and 30 September). software, or pirated movies. We found that of the victims who clicked on this kind of spam,9 50 percent were in Russia. How- nuts (240) ever, victims from many non- rofl (270) odd (240) Russian-speaking countries also omfg (270) funny (240) tweet (432) clicked on this kind of spam. As- lol (270) epic (240) suming these users did not speak Figure 3. Examples of passive spam. Note that while many spam tweets post (432) you (720) is (1,080) lmao (270) this (2,160) update (432) by (2,160) u (720) 7 awesome (240) Russian, we hypothesize that the originate from Russian domains, this strategy is also applied by spammers 2 3 4 5 6 was (1,080) blog (432) yours (720) crazy (240) 8 1 lmfao (270) SHORT_URL (2,160) content advertised in this spam outside of Russia. USER (2,160) LOL (270) entry (432) cool (240) was sufficiently enticing for vic- lolz (270) hilarious (240) tims to use translation software to haha (270) 0 so funny (240) 9 access the inappropriate content. References A Near Real-Time Detection Sys- $ (2,160) We also found that the suspended 1. C. Chen et al., “6 Million Spam tem for Suspicious URLs in Twit- rate of this type of spam by Twit- Tweets: A Large Ground Truth for ter Stream,” IEEE Trans. Dependable Figure 2. Finite-state machine-based spam template. Spammers can use this template to generate 2,160 ter is much lower than others, Timely Twitter Spam Detection,” and Secure Computing, vol. 10, no. 3, different spam tweets. because spammers have much Proc. Int’l Conf. Communications, 2015, 2013, pp. 183–195. less interaction with users, allow- pp. 7065–7070. 7. K. Thomas, “The Role of the Un- ing spammers to use this strategy 2. C. Grier et al., “@spam: The Un- derground Economy in Social group A stopped sending spam, “celebrity name” + “an eye-catching different tweets. Take the finite- successfully. derground on 140 Characters or Network Spam and Abuse,” PhD group C started sending it on 4 action” + URL. Therefore, research- state machine in Figure 2, for ex- Less,” Proc. 17th ACM Conf. Com- dissertation, Electrical Eng. and October. Groups C and E, and E ers can extract the templates and ample; it has 8 × 5 × 3 × 2 × 9 = puter and Communications Security, Computer Science Dept., Univ. of and J also displayed this type of match tweets to them to detect spam. 2,160 different routes from start to lthough researchers and in- 2010, pp. 27–37. California, Berkeley, Dec. 2013. spamming behavior. We regard We found that spammers are end. This means that spammers dustry are devoted to devel- 3. F. Benevenuto et al., “Detecting 8. C. Yang, R. Harkreader, and G. Gu, this behavior as coordinated post- now using more complex tem- can use this template to generate A oping detection and mitiga- Spammers on Twitter,” Proc. 7th “Empirical Evaluation and New ing behavior, a phenomenon in plates to generate spam. Sur- 2,160 different spam tweets with tion approaches to combat Twitter Ann. Collaboration, Electronic Messag- Design for Fighting Evolving Twit- which one group of spam tweets prisingly, spammers are using little effort. For example, spam- spam, spammers can thwart their ing, Anti-Abuse, and Spam Conf., 2010, ter Spammers,” IEEE Trans. Informa- disappears while another is being finite-state machines to generate mers can write a script that ran- efforts with ever-evolving tech- http://www.decom.ufop.br/fabricio/ tion Forensics and Security, vol. 8, no. sent. This kind of posting behavior what we have named finite-state domly chooses one option from niques, such as the three complex download/ceas10.pdf. 8, 2013, pp. 1280–1293. is more difficult to detect because machine-based spam templates each node to generate one spam spamming strategies we describe 4. J. Song, S. Lee, and J. Kim, “Spam 9. J. Oliver et al., An In-Depth Analy- spammers change the groups of (see Figure 2). One finite-state tweet. Relying on simple string here. The war with spammers is Filtering in Twitter Using Sender- sis of Abuse on Twitter, tech. report, accounts to abuse Twitter. machine has a number of states, signatures to match spam tweets becoming fiercer and is far from Receiver Relationship,” Proc. 14th Trend Micro, Sept. 2014; www. and each edge of it is denoted by will allow most of these finite- over; we should therefore continue Int’l Conf. Recent Advances in Intrusion trendmicro.com/cloud-content/us/ Finite-State Machine- one word. If we travel from the state machine-based template to analyze spammers’ behavior and Detection, 2011, pp. 301–317. pdfs/security-intelligence/white- Based Spam Template beginning to the end, we can spam tweets to escape detection. propose robust spam-detection 5. M. Egele et al., “Compa: Detecting papers/wp-an-in-depth-analysis- Some have found that most spam is have one full sentence, such as systems to make a safe Twitter en- Compromised Accounts on Social of-abuse-on-twitter.pdf. generated using specific templates,10 “lol, this tweet by you is funny + Passive Spam vironment for all users. Networks,” Proc. Ann. Network and 10. H. Gao et al., “Spam Ain’t as Di- which is logical because it is very ex- SHORT URL” in the finite-state As previously described, traditional Distributed System Security Symp., 2013, verse as It Seems: Throttling OSN pensive for spammers to write each machine. By using one finite-state spam is distributed using Twitter Acknowledgments https://www.cs.ucsb.edu/~vigna/ Spam with Templates Underneath,” tweet manually. However, the tem- machine-based spam template, functions such @ and #. However, This work is supported by ARC Linkage Proj- publications/2013_NDSS_compa.pdf. Proc. 30th Ann. Computer Security Ap- plate is often simple10—for example, spammers can generate many we also found that much spam ect LP120200266. 6. S. Lee and J. Kim, “Warningbird: plications Conf., 2014, pp. 76–85.

www.computer.org/computingedge 41 68 IT Pro March/April 2016 computer.org/ITPro 69 securing it

Chao Chen is working toward a PhD in sor at the School of Information Technol- Jonathan Oliver is a senior architect at computer science at the School of Infor- ogy, Deakin University, Australia. His Trend Micro, where he focuses on anti- mation Technology, Deakin University, research interests include network and spam and Web reputation technologies. Australia. His research interests include system security, distributed systems, and He performed postdoctoral research in network security and social network se- networking. Xiang is the chief investigator Australia and the UK, and acted as a curity. Chen received a BS in informa- of several projects in network and system data-mining consultant in Silicon Val- tion technology (with first class honors) security, funded by the Australian Re- ley. Oliver led the anti-spam R&D at from Deakin University. Contact him at search Council (ARC). He received a PhD Mailfrontier, an anti-spam start-up, [email protected]. in computer science from Deakin Univer- from 2002 to 2006. He received a PhD sity. Contact him at [email protected]. in computer science from Monash Uni- Jun Zhang is with the School of Infor- versity, Australia. Contact him at jon_ mation Technology, Deakin University, Wanlei Zhou is the chair professor of [email protected]. Australia. His research interests include IT at the School of Information Technol- network and system security, pattern ogy, Deakin University, Australia. His recognition, and multimedia processing. research interests include network secu- This article originally appeared in Zhang received a PhD from the Univer- rity, distributed and parallel systems, IT Professional, vol. 18, no. 2, 2016. sity of Wollongong, Australia. Contact bioinformatics, mobile computing, and him at [email protected]. e-learning. Zhou received a PhD from the Australian National University and Selected CS articles and Yang Xiang is the director of the Centre a DSc from Deakin University. Contact columns are available for free at for Cyber Security Research and a profes- him at [email protected]. http://ComputingNow.computer.org.

Take the CS LibraryADVERTISER INFORMATION

Advertising Personnel Southwest, California: wherever Mike Hughes Marian Anderson: Sr. Advertising Coordinator Email: [email protected] Email: [email protected] Phone: +1 805 529 6790 Phone: +1 714 816 2139 | Fax: +1 714 821 4010 Southeast: Sandy youBrown: Sr. Business Development go! Mgr. Heather Buonadies Email [email protected] Email: [email protected] Phone: +1 714 816 2144 | Fax: +1 714 821 4010 Phone: +1 973 304 4123 IEEE Computer Society magazines andFax: Transactions +1 973 585 7071 are now Advertising availableSales Representatives to subscribers (display)in the portable ePub format. Advertising Sales Representatives (Classified Line) Central, Northwest, Far East: Eric KincaidJust download the articles from the IEEE ComputerHeather Buonadies Society Digital Email: Library,[email protected] and you can read them on any deviceEmail: that [email protected] supports ePub. Phone: +1 214 673 3742 Phone: +1 973 304 4123 Fax: +1For 888 more886 8599 information, including a list of compatibleFax: +1 973 devices, 585 7071 visit

Northeast, Midwest, Europe, Middle East: Advertising Sales Representatives (Jobs Board) Ann & www.computer.org/epubDavid Schissler Email: [email protected], [email protected] Phone: +1 508 394 4026 Heather Buonadies Fax: +1 508 394 1707 Email: [email protected] Phone: +1 973 304 4123 Fax: +1 973 585 7071

42 ComputingEdge August 2016 70 IT Pro March/April 2016 Move Your Career Forward IEEE Computer Society Membership

Explore These Social Networking Resources

Build Your Knowledge

IEEE Multimedia This magazine covers multiple media types, used harmoniously to create new experiences in numerous fields including image processing and video processing.

IEEE Transactions on Mobile Computing TMC is a monthly journal that focuses on the key technical issues related to mobile computing. It publishes mature research, emphasizing issues at the link-layer and above in wireless communications, and other topics explicitly or plausibly related to mobile systems. TMC focuses on seven key technical issues: architectures, support services, algorithm/protocol design and analysis, mobile environments, mobile communication systems, applications, and emerging technologies.

IEEE Transactions on Network Science and Engineering TNSE is committed to timely publishing of peer-reviewed technical articles that deal with the theory and applications of network science and the interconnections among the elements in a system that form a network. In particular, TNSE publishes articles on understanding, prediction, and control of structures and behaviors of networks at the fundamental level.

FOR DIRECT LINKS TO THESE RESOURCES, VISIT www.computer.org/edge-resources Internet Governance Editor: Virgilio A.F. Almeida • [email protected] Whither Social Media Governance?

Virgilio A.F. Almeida • Federal University of Minas Gerais, Brazil

Danilo Doneda • Rio de Janeiro State University

Yasodara Córdova • Brazilian W3C

Although social network platforms such as Facebook, YouTube, and Twitter have their own rules, certain ethical aspects demand governance to guarantee compliance with human rights. This article explores what could impact social networks’ governance, drawing attention to possible solutions.

any social media platforms catch users’ limits and shapes for the communication among attention by delivering convenience, engage- their users calls for attention — and this is more Mment with other people, and easier access important as these platforms present themselves to content via interfaces and filters, among other as surrogates to the public space. Therefore, utilities. Consequently, much of the communica- beyond the notion of platform liability, it is man- tion that could happen in public spaces ends up datory to be clear about their potential harm to taking place in these private platforms. In this public spaces. article, we analyze the profile of social media There are fundamental differences between governance and the main issues it raises. the physical and the virtual spaces, which are According to De Nardis and Hack, social media akin to the differences between a public and a are platforms that provide “three specific techno- controlled physical space. In a physical space, logical affordances: 1. the intermediation of user- submission to the rules is usually subjected to generated content; 2. the possibility of interactivity physical presence. In the virtual environment, the among users and direct engagement with content; pervasiveness of the presence is conditioned by and 3. the ability for an individual to articulate social networking platforms’ technical attributes. network connections with other users.”1 This is In virtual spaces, users have their “fragments” the case for platforms such as Facebook, Twitter, submitted to the same rules of governance that YouTube, Yelp, Instagram, Snapchat, GitHub, and they’re exposed to when “present.” It means that others. These platforms, though, aren’t mere tech- the content, which belongs to the individual’s nical tools. When dealing with decisions regarding representation spectrum, can be everlasting and control of flows of information, platforms project spread as fragments of the presence — people elements of their intended business model. Busi- can leave a social network, but their content can ness rules aren’t those of public spaces, but rather remain as a flag, or a “piece” of the user. This of controlled spaces, where the possibilities of indi- ubiquity conferred to connected platforms is con- vidual expression and creativity are often limited ditioned using fragments in the code itself. This not only by technical constraints but by design promotes linking the user content through social choices made by the platform itself, reflecting Web spaces. For instance, Facebook’s Open Graph views from the platform controllers and developers. Protocol (see http://ogp.me/) works by centraliz- From a regulatory perspective, preserving ing the “user fragments” with linked data tech- such intermediaries from liability rules has often nologies, allowing the user to spread fragments of been seen as important to ensure free speech.1,2 information that are under Facebook’s rules, but However, their crescent tendency to prescribe virtually located in other social platforms.

82 Published by the IEEE Computer Society 1089-7801/16/$33.00 © 2016 IEEE IEEE INTERNET COMPUTING 44 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE Internet Governance Whither Social Media Governance? Editor: Virgilio A.F. Almeida • [email protected]

Factors Impacting Social of content they share and focus on. Private space Media Governance Each media type defines a set of dif- Whither Social Media Social media Social media platforms are (almost ferent patterns of interaction. These all) privately built and aim to operate interactions have a direct impact Rules of Terms Social Laws Governance? globally. Their governance is based on users’ constraints and possibili- operation of use norms on an array of elements that litera- ties. Furthermore, as the linked data ture tries to identify, such as rules of technologies advance, the Web goes Public space Virgilio A.F. Almeida • Federal University of Minas Gerais, Brazil operation or code, laws, terms of use, social and “... users can not only par- and social norms (see Figure 1). ticipate in the co-creation of value, Danilo Doneda • Rio de Janeiro State University The rules of operation compre- but also leave their own identity into Figure 1. Factors that influence social hend basically the technical factors in the system and are persuaded to keep media governance. Yasodara Córdova • Brazilian W3C which the fundamental design options using it.”3 in a given platform will find their Text, image, audio, and video have Although social network platforms such as Facebook, YouTube, and Twitter expression. It includes code but goes distinct characteristics in data pack- content that populates some task- further, encompassing also interfaces, aging, storage and availability, and oriented social networks4 might be have their own rules, certain ethical aspects demand governance to guarantee back ends, data modeling, and so on. they have different impacts on the under a completely different set of compliance with human rights. This article explores what could impact social Laws refer to laws and regulations platforms. For text, the most common governance rules than other social networks’ governance, drawing attention to possible solutions. set forth by a public authority. Regu- restrictions imposed are the amount networking platforms. Furthermore, latory frameworks can significantly of input text permitted at the format it’s worth observing the possible vary among countries, thus caus- of input boxes and the number of changes that blockchain-based sys- ing problems to a unique platform to threads allowed in nested responses. tems can bring to the rules of gov- any social media platforms catch users’ limits and shapes for the communication among adapt to all of them. Although they are The permission to post trackbacks ernance at social networking spaces. attention by delivering convenience, engage- their users calls for attention — and this is more not explicit regulations, social norms and comments also unleashes chances Platforms that don’t restrict user’s Mment with other people, and easier access important as these platforms present themselves or codes of conduct are somewhat that a platform will be used to attend expressions (such as Tumblr) are fre- to content via interfaces and filters, among other as surrogates to the public space. Therefore, important because they can constrain, social networking requisites, includ- quently accustomed to civil unrest in utilities. Consequently, much of the communica- beyond the notion of platform liability, it is man- release, or even promote the ideas of ing them in the scope of public spaces. comparison with other platforms.5 tion that could happen in public spaces ends up datory to be clear about their potential harm to communities with specific character- The question of who is responsible taking place in these private platforms. In this public spaces. istics in a social network. The socio- for comments left in pages of online Conflicts article, we analyze the profile of social media There are fundamental differences between logical concept of social norms is also news, for instance, or illegal content We can point out examples related governance and the main issues it raises. the physical and the virtual spaces, which are relevant in the sense that communities distributed in a personal blog as a to information filtering and Internet According to De Nardis and Hack, social media akin to the differences between a public and a can impose sanctions for those who link, are questions being baked while censorship, freedom of speech, and are platforms that provide “three specific techno- controlled physical space. In a physical space, don’t comply with a given behavior — the platforms become more linked privacy that may arise from the action logical affordances: 1. the intermediation of user- submission to the rules is usually subjected to and, at the other end, can induce and open to social interaction. of social media platforms on limiting generated content; 2. the possibility of interactivity physical presence. In the virtual environment, the certain types of it, even “positive” Image and video uploads are subject users’ rights that could eventually be among users and direct engagement with content; pervasiveness of the presence is conditioned by behavior. to control by sophisticated algorithms avoided in public spaces. and 3. the ability for an individual to articulate social networking platforms’ technical attributes. Furthermore, there are the terms of for copyright detection. Although this One potential cause of harm to network connections with other users.”1 This is In virtual spaces, users have their “fragments” use or, broadly speaking, the contrac- is a solution for determining revenue identity, freedom of speech, and pri- the case for platforms such as Facebook, Twitter, submitted to the same rules of governance that tual clauses that govern the platform’s on copyrighted content shared through vacy rights are the so-called “real name YouTube, Yelp, Instagram, Snapchat, GitHub, and they’re exposed to when “present.” It means that use. These are the kinds of “norms” online social networking, it’s related to policies” adopted by some social media others. These platforms, though, aren’t mere tech- the content, which belongs to the individual’s that private companies can choose to business models rather than ethics or platforms. By imposing users’ registra- nical tools. When dealing with decisions regarding representation spectrum, can be everlasting and apply to their particular relationships human rights. Textual content, notice- tion only by using the real person’s control of flows of information, platforms project spread as fragments of the presence — people with users. Usually, the main issue is ably, doesn’t receive the same atten- identification, these platforms can elements of their intended business model. Busi- can leave a social network, but their content can that the clauses in the terms of use tion. Usually, offensive content posted limit the way some people repre- ness rules aren’t those of public spaces, but rather remain as a flag, or a “piece” of the user. This often aren’t freely negotiated between in comments, for instance, can take a sent substantial traces of their per- of controlled spaces, where the possibilities of indi- ubiquity conferred to connected platforms is con- parties and tend to be overly imposed, long time to be removed. sonality and identity, as is the case vidual expression and creativity are often limited ditioned using fragments in the code itself. This due to the virtual impossibility of At GitHub, for example, there’s for some members of the lesbian, gay, not only by technical constraints but by design promotes linking the user content through social bargaining. social interaction directly at the code bisexual, and transgender (LGBT) com- choices made by the platform itself, reflecting Web spaces. For instance, Facebook’s Open Graph or in the message commits, apart munity who sometimes present them- views from the platform controllers and developers. Protocol (see http://ogp.me/) works by centraliz- Rules of Operation: from traditional spaces. The interface selves with preferred names. Using a From a regulatory perspective, preserving ing the “user fragments” with linked data tech- Technical Aspects of for searching, reading, and reacting person’s legal name is also directly such intermediaries from liability rules has often nologies, allowing the user to spread fragments of Interactions to this content is restricted to people linked to the issue of free speech, been seen as important to ensure free speech.1,2 information that are under Facebook’s rules, but The difference between social net- with skills and knowledge to operate as the mandatory association of a However, their crescent tendency to prescribe virtually located in other social platforms. works can be established by the types systems like Git or Mercurial. Hence, person’s communications with her

legal name can hurt the possibil- platforms and their use as public Understanding the Role of Persuasive User ity of anonymous speech, which is a space for expression, along with other Experiences in Committing Social Web substantial part of the legislation for activities such as payments, learning, Users,” Persuasive Technology, Springer, general free speech rights for many or unimaginable ones, a global frame- 2015, pp. 241–252. governing entities. work for governance might be desir- 6. Article 29 Data Protection Working Party, Privacy implications of social media able. Many social media issues raise Opinion 8/2014 on the on Recent Devel- platforms’ actions abound, and we can questions about the governance struc- opments on the Internet of Things, Direc- briefly mention the limits on the exer- tures and how existing structures will tive 95/46/EC of the European Parliament, cise of informational auto determina- adapt to handle new issues that are 16 Sept. 2014; http://ec.europa.eu/justice/ tion that’s often present in contractual related to the growing public impact data-protection/article-29/documenta- clauses (terms of use or privacy poli- of social media. tion/opinion-recommendation/files/2014/ cies) in which the user, in exchange for The opportunity to build global wp223_en.pdf. the use of a certain service, agrees to multistakeholder mechanisms that deal the use of his or her personal informa- with human rights, such as freedom of Virgilio A.F. Almeida is a professor in the Com- tion in ways he or she might not be expression, ethics, privacy, and data puter Science Department at the Federal comfortable with — because even nego- protection rules in global social media University of Minas Gerais (UFMG), Bra- tiation on these terms is utterly impos- should be explored. Multistakeholder zil, and currently a visiting professor at sible. Akin to these contractual choices governance bodies — involving repre- Harvard University. His research interests are several privacy choices built into sentatives of civil society, academia, include large-scale distributed systems, the the platforms’ design, such as the way private sector, and governments — Internet, social computing, and cyber poli- privacy controls (that is, commands could be a promising alternative to cies. Almeida has a PhD in computer science that can be used to choose the level of social media governance, for they have from Vanderbilt University. He’s the chair- man of the Brazilian Internet Steering Com- mittee (CGI.br). Contact him at virgilio@ It’s still extremely hard to predict all the dcc.ufmg.br.

possible conflicts that could happen in a social Danilo Doneda is a professor of civil law at the Law School of the Rio de Janeiro State network as a public space. University (UERJ). His research interests include private law and regulation, privacy, and data protection. Doneda has a dissemination desired for a bit of per- worked well for several other issues PhD in civil law from UERJ. Contact him at sonal information) are made available related to the evolution of the Internet [email protected]. to the user. Usually, privacy choices and cyberspace governance. are related to the way your personal Yasodara Córdova is a Web Specialist at the Bra- information will be shown publicly References zilian W3C office and at the Center of Stud- or to a certain class of users. Indeed, 1. L. DeNardis and A.M. Hackl, “Internet Gov- ies for Web Technologies. Her work focuses more sophisticated privacy controls are ernance by Social Media Platforms,” Tele- on researching and applying the technologi- actually being discussed, as the notion comm. Policy, vol. 39, no. 9, 2015. cal aspects of the Web in the social field. of the individual exercising “granular 2. M. Mueller, “Hyper-Transparency and Córdova has an MBA in strategic manage- consent” over his or her personal infor- Social Control: Social Media as Magnets for ment. Contact her at [email protected]. mation in a given platform is present in Regulation,” Telecomm. Policy, vol. 39, no. the document of recent developments 9, 2015; doi:10.1016/j.telpol.2015.05.001. for the Internet of Things.6 3. M. Oermann, M. Lose, and J.-H. Schmidt, “Approaching Social Media Governance,” This article originally appeared in HIIG Discussion Paper Series, discussion IEEE Internet Computing, vol. 20, t’s still extremely hard to predict paper no. 2014-05. no. 2, 2016 I all the possible conflicts that could 4. J. Xu et al., “Civil Unrest Prediction: A happen in a social network as a public Tumblr-Based Exploration,” Social Com- space. As platforms continue to evolve puting, Behavioral-Cultural Modeling and globally, new issues related to gov- Prediction, Springer, 2014, pp. 403–411. Selected CS articles and columns ernance may arise. With the increase 5. Oduor, Michael, and Harri Oinas-Kukkonen. are also available for free at http:// of convergence between social media “A System’s Self-referential Persuasion: ComputingNow.computer.org.

84 www.computer.org/internet/ IEEE INTERNET COMPUTING 46 ComputingEdge August 2016 THE PRAGMATIC Editor: Eoin Woods Endava ARCHITECT [email protected]

Harnessing the Power of Architectural Design Principles

Eoin Woods

PEOPLE OFTEN TALK about a sys- sonably de ne a software design ample from the banking domain. In tem’s “architectural design princi- principle as short, this principle ples” (or just “architecture principles”). But without a clear de nition a fundamental truth or proposi- • provides a clear name that hints of “principle,” it’s not always obvi- tion serving as the foundation for at how we’re meant to respond; ous what this means. By de ning the action with regard to deciding on a • offers a brief, clear description; role and bene ts of architectural de- software system’s workings. • makes the driver behind the sign principles, we can capitalize on principle clear; a very useful technique. The key point is that a principle is a • de nes its domain of applicabil- In this column, I de ne architec- clear statement of intent that guides ity (messaging for certain data tural design principles, explore what our design work. types for applications classi ed good principles look like, and de- Other de nitions of design prin- in certain ways—other people scribe when using principles in archi- ciples have been proposed. In the needn’t worry about this prin- tectural practice might be valuable. book Architecture Principles,1 ciple); and Danny Greefhorst and Erik Proper • explains why this principle is De ning Principles created probably the most compre- important so that we can decide I’m deliberately not making a hard hensive de nition: when to apply it. distinction between design and architecture because I don’t think it a declarative statement that nor- This input into the design process is adds anything useful in this context. matively prescribes a property of useful because it emphasizes the orga- For the purposes of this discussion, the design of an artefact, which is nization’s priorities, explaining when I’m talking about design principles necessary to ensure that the artefact and why the principle is important. that apply equally well at both more meets its essential requirements. detailed and more abstract levels. The Bene ts The Oxford English Dictionary Although this de nition is a little of Architecture Principles states that a principle is “a funda- abstract, it clari es the design prin- Architecture principles epitomize ar- mental truth or proposition serving ciple’s role as ensuring that some as- chitecture’s function: to clearly de- as the foundation for belief or ac- pect of your architecture meets some ne the necessary constraints on a tion,” and that a design is a “plan or aspect of its requirements. system’s design without prescriptively drawing produced to show the look But enough about abstract de ni- de ning all the design details. In a and function or workings of a build- tions. What does a real architecture previous column, I said that architec- ing, garment, or other object before principle look like? Figure 1 shows a ture deals with the “gaps” between it is built or made.” So, we can rea- simple principle based on a real ex- things.2 Thus, architecture principles

Name: Prefer standardized messaging Description: When deciding on message formats between applications: 1. We prefer to use industry standard formats (such as FpML, FIX, and ISO 20022) with well-documented in-house extensions where required. 2. Where this isn’t possible, we aim to use an established organizational standard (such as EQMF). 3. Where neither is possible, a well-dened, published application-specic format should be used.

Driver: Minimize interapplication coupling and maximize interoperability while avoiding excessive development costs.

Applicability: This principle applies to all messaging for business transactions or reference data for applications with the disposition “strategic,” “maintain,” or “improve.” Rationale: • Third-party formats are well thought out, standardize our workows with the industry, make interoperation with third-party software and services easier, and are often handled via standard third-party libraries. • Established group standards align business practice and conventions across the group; are familiar to developers, administrators, and testers; and often have established libraries or processing patterns. • Where a local message format must be used, it should be treated like a local standard to maximize reuse and minimize interoperability surprises. Publishing the format keeps us honest and forces us to dene it properly. • All these options might be initially more expensive than an ad hoc point-to-point message format. But, in the long run, we know that ad hoc formats are extremely expensive.

FIGURE 1. An architecture principle based on a real example from the banking domain.

help establish boundaries and priori- systems must be suitable for high- and short on actionable specifics. It’s ties without micromanaging how ev- availability deployment might justify difficult to produce a set of princi- eryone performs their work. A good building multinode operation capa- ples that people find valuable. set of architecture principles offers bilities into all systems, even if this My long-time collaborator Nick the following key benefits. isn’t the cheapest option for the im- Rozanski has identified a good ar- mediate future. chitecture principle’s essential crite- Context for Design Decisions ria,3 which are summarized in Table Principles can clarify priorities and Enhanced Collaboration, 1. These characteristics take effort constraints, helping people make Communication, and Shared Values to achieve but result in principles consistent, informed design deci- Like many architecture artifacts, that are much more likely to be valu- sions. In fact, I’ve found they can principles must be developed by able and provide significant, action- make abstract ideas such as business groups, not individuals. This ensures able guidance. goals more accessible and help de- that they’re validated early and that signers make technical decisions that the group feels collective ownership When to Violate a Principle support them. I’ll return to this idea of them. Moreover, this helps people Principles exist to guide the design in more depth in a future column. collaborate and build shared values, process and aid consistency. But ar- fostering a mutual understanding of chitecture principles will occasion- Justification for Decisions, what is and isn’t important. ally be broken—sometimes because Cost, and Time people don’t realize they exist (per- As designers, we often face situa- Defining Good haps there are too many of them) tions in which the right plan costs Architecture Principles and sometimes because people ig- more or takes longer than we’d like, The practical problem with prin- nore them (in which case, you must but it’s difficult to explain clearly ciples is that they’re hard to define find out why). and succinctly why it’s the right well. It’s easy to produce lists of self- Breaking a principle comes with a plan. A set of clear principles can evident statements or rambling state- cost, often a long-term one. So when provide a basis for that explanation. ments of intent that no one can apply someone violates a principle, it must For example, the principle that all because they’re long on philosophy be for a justifiable reason. When this

4816 IEEE SOFTWAREComputingEdge | WWW.COMPUTER.ORG/SOFTWARE | @IEEESOFTWARE August 2016 THE PRAGMATIC ARCHITECT THE PRAGMATIC ARCHITECT

This article originally appeared in IEEE Software, vol. 33, no. 4, 2016.

Name: Prefer standardized messaging Characteristics of a good architecture principle. Description: When deciding on message formats between applications: Characteristic Description 1. We prefer to use industry standard formats (such as FpML, FIX, and ISO 20022) with well-documented in-house extensions where required. Constructive The principle has a definite purpose, provides specific guidance, and guides decision making.

2. Where this isn’t possible, we aim to use an established organizational standard (such as EQMF). 1 TABLE 3. Where neither is possible, a well-dened, published application-specic format should be used. Reasoned The principle is strongly motivated by business drivers, goals, and other principles. It’s rational, logical, and consistent.

Driver: Minimize interapplication coupling and maximize interoperability while avoiding excessive development costs. Well-articulated The principle is clearly written so that all the necessary stakeholders comprehend it. Applicability: This principle applies to all messaging for business transactions or reference data for applications with the disposition “strategic,” “maintain,” or “improve.” Testable Valuable principles are long lived. To be useful, a principle must be capable of being assessed objectively, to see whether it’s been adhered to and, if not, where the exceptions are. Rationale: • Third-party formats are well thought out, standardize our workows with the industry, make interoperation with third-party software and services easier, and are often handled via standard third-party libraries. Significant A self-evident principle is rarely of value. A principle shouldn’t just be a truism. If the opposite statement could ever be • Established group standards align business practice and conventions across the group; are familiar to developers, true, then the principle probably isn’t very valuable. administrators, and testers; and often have established libraries or processing patterns. • Where a local message format must be used, it should be treated like a local standard to maximize reuse and minimize interoperability surprises. Publishing the format keeps us honest and forces us to dene it properly. • All these options might be initially more expensive than an ad hoc point-to-point message format. But, in the long run, we know that ad hoc formats are extremely expensive. occurs, you must check whether the tecture Principles: The Cornerstones 3. N. Rozanski and E. Woods, Soft- benefits of breaking the principle of Enterprise Architecture, Springer, ware Systems Architecture, 2nd ed., outweigh the costs. 2011. Addison- Wesley, 2011. FIGURE 1. An architecture principle based on a real example from the banking domain. Broken design principles can also 2. E. Woods, “Architecting in the Gaps: provide valuable information for ar- A Metaphor for Architecture Work,” EOIN WOODS is the chief technology of- chitects. First, the rationale for break- IEEE Software, vol. 32, no. 4, 2015, ficer at Endava. Contact him at eoin.woods@ help establish boundaries and priori- systems must be suitable for high- and short on actionable specifics. It’s ing the principle highlights some im- pp. 33–35. endava.com. ties without micromanaging how ev- availability deployment might justify difficult to produce a set of princi- portant aspect of the system design. eryone performs their work. A good building multinode operation capa- ples that people find valuable. Second, a principle that’s justifiably set of architecture principles offers bilities into all systems, even if this My long-time collaborator Nick and routinely broken signals a mis- the following key benefits. isn’t the cheapest option for the im- Rozanski has identified a good ar- match between assumptions and re- mediate future. chitecture principle’s essential crite- ality and must be changed. Finally, Context for Design Decisions ria,3 which are summarized in Table knowing where in the system the de- Principles can clarify priorities and Enhanced Collaboration, 1. These characteristics take effort sign principles are being violated will On Computing constraints, helping people make Communication, and Shared Values to achieve but result in principles help you deal with these nonstandard consistent, informed design deci- Like many architecture artifacts, that are much more likely to be valu- aspects as the system evolves. podcast sions. In fact, I’ve found they can principles must be developed by able and provide significant, action- www.computer.org/oncomputing make abstract ideas such as business groups, not individuals. This ensures able guidance. COMPUTER ENTREPRENEUR goals more accessible and help de- that they’re validated early and that lthough people often talk AWARD signers make technical decisions that the group feels collective ownership When to Violate a Principle about their systems’ archi- support them. I’ll return to this idea of them. Moreover, this helps people Principles exist to guide the design tectural principles, they’re In 1982, on the occasion of its All members of the profession A thirtieth anniversary, the IEEE are invited to nominate a in more depth in a future column. collaborate and build shared values, process and aid consistency. But ar- often hard-pressed to name them and Computer Society established colleague who they consider fostering a mutual understanding of chitecture principles will occasion- explain their rationale. A little time the Computer Entrepreneur most eligible to be considered Award to recognize and for this award. Awarded Justification for Decisions, what is and isn’t important. ally be broken—sometimes because spent during a system’s life cycle, par- honor the technical managers to individuals whose Cost, and Time people don’t realize they exist (per- ticularly early in its development, to and entrepreneurial leaders entrepreneurial leadership is As designers, we often face situa- Defining Good haps there are too many of them) identify, debate, capture, and com- who are responsible for the responsible for the growth of growth of some segment of the some segment of the computer tions in which the right plan costs Architecture Principles and sometimes because people ig- municate a coherent set of design computer industry. The efforts industry. more or takes longer than we’d like, The practical problem with prin- nore them (in which case, you must principles can be very valuable. It’s must have taken place over DEADLINE FOR 2017 AWARD fifteen years earlier, and but it’s difficult to explain clearly ciples is that they’re hard to define find out why). not an easy process, but the end result NOMINATIONS and succinctly why it’s the right well. It’s easy to produce lists of self- Breaking a principle comes with a helps align theory and practice. the industry effects must be generally and openly visible. DUE: 15 OCTOBER 2016 plan. A set of clear principles can evident statements or rambling state- cost, often a long-term one. So when provide a basis for that explanation. ments of intent that no one can apply someone violates a principle, it must References AWARD SITE: https://www.computer.org/web/awards/entrepreneur For example, the principle that all because they’re long on philosophy be for a justifiable reason. When this 1. D. Greefhorst and E. Proper, Archi- www.computer.org/awards

16 IEEE SOFTWARE | WWW.COMPUTER.ORG/SOFTWARE | @IEEESOFTWARE www.computer.org/computingedge JULY/AUGUST 2016 | IEEE SOFTWARE 17 49 Experience the Newest and Most Advanced Thinking in Big Data Analytics

03 November 2016 | Austin, TX

Big Data: Big Hype or Big Imperative? Rock Star Speakers BOTH. Business departments know the promise of big data―and they want it! But newly minted data scientists can’t yet meet expectations, and technologies remain immature. Yes, big data is transforming the way we do―everything. But knowing that doesn’t help you decide what steps to take tomorrow to assure your company’s future.

That’s why May 24 is your real-world answer. Kirk Borne Satyam Priyadarshy Bill Franks Come meet the experts who are grappling with and Principal Data Chief Data Scientist, Chief Analytics solving the problems you face in mining the value Scientist, Halliburton Offi cer, Teradata of big data. You literally can’t aﬀ ord to miss the all Booz Allen Hamilton new Rock Stars of Big Data 2016.

www.computer.org/bda FROM THE EDITORS

Trust Me. Trust Me Not.

give my house keys to my neighbor to feed both trusted computing and trust manage- I my cat and collect my mail while I’m on ment systems are technologies that rely heav- vacation. I’m not concerned that she’ll take ily on cryptography to enforce policy. anything of value, peruse my personal belong- Perhaps, instead of viewing (dis)trust as a ings, or intentionally do any damage. I don’t problem that needs to be fixed by security, we put any controls in place to try to prevent her should embrace trust as a complement to secu- from doing these things. rity. Is it possible to rely on trust as a compo- I give my credit card to a waiter in a restau- nent in a system? How we can improve people’s rant to pay the bill. He disappears with it for trust in systems? To what extent can we coun- a few minutes to run it through a credit card ter problems that arise because people mistak- reader. He could be making a copy of the card, enly place trust in systems they shouldn’t? buying things online, or stealing from me in The organizational psychology commu- any number of ways. It happens. Fortunately, nity has studied trust extensively. In their there are liability limits so that even if he did, seminal paper, “An Integrative Model of Orga- I would be out at most 50 bucks. But, I never nizational Trust,” Roger C. Mayer and his actually worry about it. colleagues give a definition that matches our Bill Horne I give my expensive camera to a stranger colloquial notion of trust.3 It’s worth under- Associate Editor in Chief to take a picture of me and my wife at a tour- standing because it can give us insight into the ist location. She happily does so, and we trust relationships between actors in the secu- might even exchange a few pleasantries. For a rity ecosystem and can help explain why some moment we are friends. She could easily have are trusted and some aren’t. run away with my camera, or smashed it on In this model, trust is a relationship the ground. between two parties: the trustor, which is We trust people and organizations all the the entity that is trusting something, and time, often with little evidence that they’re the trustee, which is the entity being trusted. trustworthy. Noted author Stephen Covey said, These parties could be individuals, but they “Trust is the glue of life. It’s the most essential could also be organizations. ingredient in effective communication. It’s the The degree to which the trustor trusts the foundational principle that holds all relation- trustee first depends on the willingness of ships.”1 Arguably, it’s hard to imagine how we the trustor to trust. Some people might have could function in a world without trust. an inherent propensity to trust, while others Trust and security are clearly related. might be more reluctant. Cultural factors can In the colloquial setting, like the examples have a great influence here. above, the more I trust someone or some- But the more interesting and complex part thing, the fewer security controls I apparently of the relationship is with the trustee. There need to protect myself. Indeed, one definition are three trustee characteristics that enable of trust is “the willingness of a party to be vul- trust: ability, integrity, and benevolence. nerable to the actions of another party based Ability refers to the trustee’s competence on the expectation that the other will perform or perceived expertise. If the trustee is being a particular action important to the trustor, relied on to perform some action on the trus- irrespective of the ability to monitor or con- tor’s behalf, then it should be capable of exe- trol that other party.”2 cuting the task. One of the reasons I trust my But in the world of security, we tend to use doctors is that I believe their medical expertise the word “trust” quite differently. In security, will enable them to provide me with the med- we typically assume that parties distrust each ical care I need. Conversely, we’re unlikely to other by default, and then we show how secu- trust someone we perceive as incompetent. rity can be an enabling technology for those Integrity refers to the values that the parties to overcome this barrier. For example, trustee demonstrates. If I believe someone

2469-7087/16/$33.00 © 2016 IEEE Published by the IEEE Computer Society August 2016 51 1540-7993/16/$33.00 © 2016 IEEE Copublished by the IEEE Computer and Reliability Societies May/June 2016 3 FROM THE EDITORS

adheres to a set of principles and might trust them with my cat, but this in an unsafe neighborhood, or has a sense of justice, then I’m not with other things. in a crowded, dark location. more likely to trust this person with Why do I trust the waiter? In What does this have to do with something important. If someone the US, we’re used to a waiter dis- security? has a demonstrated history of trust- appearing with our credit card, but First, there’s a relationship worthy behavior, then I’m likely to in other countries, the waiter often between trust and risk. We know we believe this person will continue to comes to the table with a credit can’t achieve 100 percent security, so behave that way in the future. Peo- card–reading machine. In those we have to make cost–benefit deci- ple who are perceived to be open countries, patrons might be used sions based on risk. Evaluating the about their thoughts and likelihood of a particular feelings are generally more Perhaps, instead of viewing (dis)trust risk is difficult, but our trustworthy than those calculation of that likeli- who don’t share informa- as a problem that needs to be fixed hood could easily include tion so freely. Finally, a by security, we should embrace trust trust. For example, we third party recommending might view the likeli- someone as trustworthy as a complement to security. hood of certain insider can carry great weight. threats as low because we Benevolence is the trust our employees. Or extent to which the trus- we might perceive risks tor believes that the trustee wants to keeping their card with them at associated with third parties as low to do good to the trustor. Although all times and would be suspicious because we have trusted relation- it plays a clear role in trusting indi- if a waiter disappeared with it. ships with them. The more I trust viduals, this isn’t likely to be a fac- Those people’s inherent propensity someone, the lower my evaluation tor in interorganizational trust.4 to trust waiters might be different of those risk likelihoods. Might it be Corporations are generally under- from an American’s. Probably the possible to lower risk by increasing stood to be acting to improve their main reason I trust waiters is that trust? Perhaps an effective means to firms’ profits, although they might I’ve given my credit card to them address insider threat would be to engage in some benevolent activi- hundreds of times, and I’ve never improve the trust relationships with ties, like philanthropy. had a problem. Waiters in the aggre- our employees. Let’s revisit the examples at the gate have a history of trustworthy Next, I’ve become fascinated start of this article. behavior. This particular restaurant (and somewhat disappointed) by Why do I trust my neighbor? has a good reputation (third-party the security industry’s reliance on Ability isn’t much of an issue here; recommendation). But I might be fear, uncertainty, and doubt as a she is certainly capable of feeding hesitant to hand over my card if I marketing tool. Can’t we do better? my cat. In fact, she is a cat owner were eating at some sketchy place In the context of trust, the reason herself, so she has some expertise in in a bad neighborhood, because for this tactic is clear: because the cat care. As for integrity, I’ve spent I might associate it with other adversaries have such strong charac- enough time with her to know that untrustworthy activities. teristics of distrust. They’re gener- she’s a good person, a caring mother, Why do I trust strangers with ally considered highly capable, yet and involved in the community. my camera? I might have to give their values and integrity are polar Because of this history of trustwor- them a quick lesson in how to use opposite of those they’re attacking. thy behavior, I believe she would my camera, but it’s easy, so I don’t Few abide by the rule of law and, never do anything illegal or unethi- doubt they can do it. The interest- if anything, they seek to do harm cal. Everybody I know thinks she’s ing thing is that they give subtle sig- rather than good. terrific (third-party recommenda- nals about their integrity. They look In contrast, the security indus- tion). And she’s more than happy like other people I trust. Perhaps try has a much harder time mak- to do this favor for me; she’s truly they’re behaving in a good-natured ing a case for trust. For example, benevolent. Although I am com- way. Like waiters, strangers (as a economist Ian Grigg argues that “in fortable with her, if instead I hired class of people) have demonstrated the market for security goods, buy- someone I didn’t know, I might hide a history of trustworthy behavior ers are not informed sufficiently my valuables or lock some doors. with respect to taking these kinds of to make rational decisions.” And While this person might have the photos. There are other subtle clues as a result, “decisions are therefore ability to feed the cat, I haven’t been that might depend on the context. I made primarily on the basis of exo- able to evaluate their integrity. So I might be much more hesitant to do genous factors that do not strongly

52 ComputingEdge August 2016 4 IEEE Security & Privacy May/June 2016 FROM THE EDITORS

PURPOSE: The IEEE Computer Society is the world’s largest association of computing professionals and is the leading provider of technical information in the field. Visit our adheres to a set of principles and might trust them with my cat, but this in an unsafe neighborhood, or (causally) relate to the nominal been compromised and is under website at www.computer.org. has a sense of justice, then I’m not with other things. in a crowded, dark location. security goal of the good.” 5 In short, an adversary’s control. Machines OMBUDSMAN: Email ombudsman@ computer.org. more likely to trust this person with Why do I trust the waiter? In What does this have to do with it’s hard for the buyer to evaluate the don’t have values. They can’t be Next Board Meeting: 13–14 November 2016, something important. If someone the US, we’re used to a waiter dis- security? vendor’s ability. benevolent. Perhaps we do have it New Brunswick, NJ, USA has a demonstrated history of trust- appearing with our credit card, but First, there’s a relationship Industry analyst reports carry right in security. Perhaps trust can EXECUTIVE COMMITTEE worthy behavior, then I’m likely to in other countries, the waiter often between trust and risk. We know we weight because they are essentially only be achieved by the annihila- President: Roger U. Fujii believe this person will continue to comes to the table with a credit can’t achieve 100 percent security, so third-party recommendations. But tion of distrust. President-Elect: Jean-Luc Gaudiot; Past President: Thomas M. Conte; Secretary: Gregory behave that way in the future. Peo- card–reading machine. In those we have to make cost–benefit deci- what more can we do? Perhaps the ExecutiveT. Byrd; Committee Treasurer: (ExCom) Forrest Members:Shull; VP, Member Christian ple who are perceived to be open countries, patrons might be used sions based on risk. Evaluating the security industry could rely less on Hansen,& President; Geographic Dennis Activities: Hoffman, Nita K. Patel;Jr. Past V P,President; Publications: David S. Ebert; VP, Professional about their thoughts and likelihood of a particular fear, uncertainty, and doubt as a mar- n the physical world, our day-to- Jeffrey &Voas, Educational Sr. Past Activities:President; AndyW. Eric T. Chen;Wong, V P,VP feelings are generally more Perhaps, instead of viewing (dis)trust risk is difficult, but our keting message if we figured out ways I day lives are arguably governed Publications;Standards Carole Activities: Graas, MarkVP Meetings Paulk; VP, and Technical Conferences; trustworthy than those calculation of that likeli- to send stronger messages to our cus- more by trust than security. As tech- & Conference Activities: Hausi A. Müller; 2016 Joe Childs,IEEE VP Director Membership; & Delegate Shiuhpyng Division VIII:Winston John Shieh,W. who don’t share informa- as a problem that needs to be fixed hood could easily include tomers about our ability, integrity, nology continues to become more VP TechnicalWalz; 2016Activities; IEEE Director Scott Abrams, & Delegate Secretary; Division Robert V: tion so freely. Finally, a trust. For example, we and benevolence characteristics. engrained in our physical lives, is it Harold Javid; 2017 IEEE Director-Elect & Delegate by security, we should embrace trust Loomis,Division Treasurer; V: Dejan Pradeep S. Miloji Ramuhalli,čić Secretary third party recommending might view the likeli- For a more technical exam- possible that trust gives way to secu- BOARD OF GOVERNORS someone as trustworthy as a complement to security. hood of certain insider ple, consider fully homomorphic rity? What’s the future of trust? AdministrativeTerm Expriring Committee 2016: David (AdCom) A. Bader, Members: Pierre can carry great weight. threats as low because we encryption, which is a remarkable Scott Abrams,Bourque, Evelyn Dennis H. J. Hirt,Frailey, Charles Jill I. Gostin,H. Recchia, Atsuhiro Jason W. Benevolence is the trust our employees. Or mathematical advancement that, References Goto, Rob Reilly, Christina M. Schober Rupe, AlfredTerm ExpiringM. Stevens, 2017: and David Jeffrey Lomet, Voas Ming C. Lin, extent to which the trus- we might perceive risks once it becomes practical, will 1. S.R. Covey, The 7 Habits of Highly Gregory T. Byrd, Alfredo Benso, Forrest Shull, tor believes that the trustee wants to keeping their card with them at associated with third parties as low enable all sorts of interesting things. Effective People: Restoring the Char- Fabrizio Lombardi, Hausi A. Müller http://rs.ieee.orgTerm Expiring 2018: Ann DeMarle, Fred Douglis, to do good to the trustor. Although all times and would be suspicious because we have trusted relation- But often, the underlying narrative acter Ethic, Free Press, 2004. Vladimir Getov, Bruce M. McMillin, Cecilia Metra, it plays a clear role in trusting indi- if a waiter disappeared with it. ships with them. The more I trust is that it will enable computation in 2. H.W. Kee and R.E. Knox, “Concep- Kunio Uchiyama, Stefano Zanero viduals, this isn’t likely to be a fac- Those people’s inherent propensity someone, the lower my evaluation the cloud with an untrusted cloud tual and Methodological Consid- TheEXECUTIVE IEEE Reliability STAFF Society (RS) is a technical society tor in interorganizational trust.4 to trust waiters might be different of those risk likelihoods. Might it be service provider. But is it even desir- erations in the Study of Trust and within theExecutive IEEE, which Director: is the Angela world’s R. leading Burgess; professional Director, associationGovernance for the advancement & Associate Executive of technology. Director: The Anne RS is Corporations are generally under- from an American’s. Probably the possible to lower risk by increasing able to have a business relationship Suspicion,” J. Conflict Resolution, engagedMarie in the Kelly; engineering Director, disciplines Finance & of Accounting: hardware, software, stood to be acting to improve their main reason I trust waiters is that trust? Perhaps an effective means to with an untrusted service provider, 1970, vol. 14, no. 3, pp. 357–366. and humanSunny factors. Hwang; Its Director, focus on Information the broad aspects Technology of reliability & Services: Sumit Kacker; Director, Membership firms’ profits, although they might I’ve given my credit card to them address insider threat would be to especially in business-to-business 3. R.C. Mayer, J.H. Davis, and F.D. allows theDevelopment: RS to be seen Eric as Berkowitz; the IEEE Specialty Director, Engineering Products engage in some benevolent activi- hundreds of times, and I’ve never improve the trust relationships with transactions? Is there enough value Schoorman, “An Integrative Model organization.& Services: The IEEE Evan ReliabilityM. Butterfield; Society Director, is concerned Sales & with attainingMarketing: and sustaining Chris theseJensen design attributes throughout ties, like philanthropy. had a problem. Waiters in the aggre- our employees. in having a trusted relationship with of Organizational Trust,” Academy the total life cycle. The Reliability Society has the Let’s revisit the examples at the gate have a history of trustworthy Next, I’ve become fascinated a service provider that outweighs of Management Rev., vol. 20, no. 3, COMPUTER SOCIETY OFFICES management,Washington, resources, D.C.: 2001and administrativeL St., Ste. 700, and technical start of this article. behavior. This particular restaurant (and somewhat disappointed) by the possible security risks? For 1995, pp. 709–734. structuresWashington, to develop D.C. and 20036-4928 to provide technical information Why do I trust my neighbor? has a good reputation (third-party the security industry’s reliance on example, I can build on trusted rela- 4. F.D. Schoorman, R.C. Mayer, and via publications,Phone: +1 training, 202 371 0101 conferences, • Fax: +1 202and 728 technical 9614 library (IEEEEmail: Xplore) [email protected] data to its members and the Specialty Ability isn’t much of an issue here; recommendation). But I might be fear, uncertainty, and doubt as a tionships to help me build and grow J.H. Davis, “An Integrative Model of Los Alamitos: 10662 Los Vaqueros Circle, Los she is certainly capable of feeding hesitant to hand over my card if I marketing tool. Can’t we do better? my business. Organizational Trust: Past, Present EngineeringAlamitos, community. CA 90720 The • Phone: IEEE +1 Reliability 714 821 8380Society • has 28 chaptersEmail: and [email protected] in 60 countries worldwide. my cat. In fact, she is a cat owner were eating at some sketchy place In the context of trust, the reason Trust also plays a role in under- and Future,” Academy of Manage- TheMembership Reliability &Society Publication is the OrdersIEEE professional society herself, so she has some expertise in in a bad neighborhood, because for this tactic is clear: because the standing why users engage in risky ment Rev., vol. 32, no. 2, 2007, pp. for ReliabilityPhone: Engineering,+1 800 272 6657 along • Fax: with +1 other 714 821 Specialty 4641 • Email: [email protected] Engineering disciplines. These disciplines are design cat care. As for integrity, I’ve spent I might associate it with other adversaries have such strong charac- behaviors like opening email attach- 344–354. Asia/Pacific: Watanabe Building, 1-4-2 Minami- enough time with her to know that untrustworthy activities. teristics of distrust. They’re gener- ments, installing apps on their phone, 5. I. Grigg, “The Market for Silver engineeringAoyama, fields Minato-ku, that apply Tokyo scientific 107-0062, knowledge Japan • so that she’s a good person, a caring mother, Why do I trust strangers with ally considered highly capable, yet or being subject to social engineer- Bullets,” 2 Mar. 2008; www.iang their specificPhone: +81attributes 3 3408 are3118 designed • Fax: +81 into 3 3408the system3553 • / productEmail: / device [email protected] / process to assure that it will perform its and involved in the community. my camera? I might have to give their values and integrity are polar ing. Can the psychology of trust help .org/papers/market_for_silver intendedIEEE function BOARD for OFthe DIRECTORS required duration within a given Because of this history of trustwor- them a quick lesson in how to use opposite of those they’re attacking. to mitigate these behaviors? How do _bullets.pdf. environment,President including & CEO: Barrythe ability L. Shoop; to test President- and support thy behavior, I believe she would my camera, but it’s easy, so I don’t Few abide by the rule of law and, we help users distinguish between 6. S.E. Schechter et al., “The Emper- it throughoutElect: Karen its total Bartleson; life cycle. Past This President is accomplished: Howard E. Michel; Secretary: Parviz Famouri; Treasurer: never do anything illegal or unethi- doubt they can do it. The interest- if anything, they seek to do harm trustworthy and untrustworthy or’s New Security Indicators,” Proc. concurrently with other design disciplines by contributing to the planningJerry L. Hudgins; and selection Director of & the President, system IEEE-architecture, cal. Everybody I know thinks she’s ing thing is that they give subtle sig- rather than good. sources? Unfortunately, most of the IEEE Symp. Security and Privacy, USA: Peter Alan Eckstein; Director & President, design implementation, materials, processes, and 2007, pp. 51–65. Standards Association: Bruce P. Kraemer; Director terrific (third-party recommenda- nals about their integrity. They look In contrast, the security indus- efforts in this direction have had components;& VP, Educational followed by Activities: verifying S.K. the Ramesh;selections made by tion). And she’s more than happy like other people I trust. Perhaps try has a much harder time mak- limited success. For example, much thoroughDirector analysis & VP, and Membership test and then and sustainment. Geographic to do this favor for me; she’s truly they’re behaving in a good-natured ing a case for trust. For example, work has focused on trying to estab- VisitActivities: the IEEE Wai-ChoongReliability Society (Lawrence) website Wong; as it is the Director & VP, Publication Services and Products: gateway to the many resources that the RS makes available benevolent. Although I am com- way. Like waiters, strangers (as a economist Ian Grigg argues that “in lish trusted identities through cer- This article originally appeared in Sheila Hemami; Director & VP, Technical fortable with her, if instead I hired class of people) have demonstrated the market for security goods, buy- tificate authorities. But it’s unclear IEEE Security & Privacy, vol. 14, to its membersActivities: and Jose others M.F. Moura; interested Director in the & Delegate broad aspects of ReliabilityDivision and Specialty V: Harold Engineering. Javid; Director & Delegate someone I didn’t know, I might hide a history of trustworthy behavior ers are not informed sufficiently whether most users really pay atten- no. 3, 2016. Division VIII: John W. Walz my valuables or lock some doors. with respect to taking these kinds of to make rational decisions.” And tion to these things.6 While this person might have the photos. There are other subtle clues as a result, “decisions are therefore On the other hand, it’s hard Selected CS articles and columns ability to feed the cat, I haven’t been that might depend on the context. I made primarily on the basis of exo- to trust a machine. It’s too dif- are also available for free at revised 10 June 2016 able to evaluate their integrity. So I might be much more hesitant to do genous factors that do not strongly ficult to know if a machine has http://ComputingNow.computer.org.

www.computer.org/computingedge 53 4 IEEE Security & Privacy May/June 2016 www.computer.org/security 5 COMPUTING CAREERS

Finding the Social-Networking Job You Want

or this issue of ComputingEdge, we and everything in between. There will be a need asked Michael Tsikerdekis—an assistant for people who can design algorithms that accom- F professor of information and commu- modate large datasets and operations on them. nications technology at the University of Ken- There will also be a demand for software engi- tucky—about career opportunities in the fi eld of neers, database administrators and designers, social networking. Tsikerdekis’ research interests and even online community organizers who can are social engineering, data mining, collabora- support dynamic environments like those found in tion technologies, and social-media design and social networking. development. He coauthored “Detecting and Pre- The analytics of social-networking data will venting Online Identity Deception in Social Net- also become very important. Some companies— working Services” from IEEE Internet Computing’s such as Facebook—already have departments May–June 2015 issue. that collect and study social-networking data for various purposes. Nonetheless, we will also see ComputingEdge: What social networking-related businesses established solely to collect and study careers will see the most growth in the next sev- other companies’ data. Thus, graduates inter- eral years? ested in analytics are likely to see a big demand for their skills. Tsikerdekis: We believe that social networking as a fi eld and market will grow in two ways over the ComputingEdge: What would you tell college next several years. First, there is still a need for students to give them an advantage over the social-media-centered products—desktop, mobile, competition?

54 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE Tsikerdekis: The short answer is: attempt to gain LinkedIn profi le. Beyond that, maintaining the knowledge outside your core specialization. For existing channels of communication is important. industry, the most important characteristic in This doesn’t mean bombarding people with programmers and other job candidates appears emails but rather it means keeping the lines of to be a variety of skills rather than an explicit communication open. If the six-degrees-of-sep- specialization. Successful social-media platforms aration theory is true, you are only four or fi ve employ a holistic approach to designing their people away from virtually any place you could networks. imagine working for. With the explosion of start-up companies, many coders will play a crucial role in their com- ComputingEdge: What should applicants keep panies’ future and will help form their employer’s in mind when applying for social networking identity. Therefore, they will need a clear vision jobs? and a broad understanding of the fi eld. We need our coders to be communicators, leaders, design- Tsikerdekis: Social-networking positions are ers, and analysts. Programmers aren’t likely to careers, not just jobs. Not every company will give master all of these areas but instead should focus you space to grow and climb up the ladder in the on one area and get a general knowledge of oth- traditional sense. Many social-networking com- ers. Among other benefi ts, this will help them panies are small, yet sustainable and successful appreciate fellow team members’ points of view. in what they do. They off er their employees the Also, good communication skills will help stu- potential for growth or perhaps a stepping stone dents integrate well into companies that hire them. to something else. I would suggest doing your due diligence when looking for jobs but don’t ComputingEdge: What advice would you give dismiss those that seem small and insignifi cant. people changing careers midstream? Social networking is a fast, dynamic, and innovative industry willing to explore and grow. Potential Tsikerdekis: If they are entering the fi eld of hires should do the same. social networking, they are likely to encounter a diverse and ever-changing competitive environment. They will have to compete with new college omputingEdge’s Lori Cameron inter- graduates who may appear—at least to employ- viewed Tsikerdekis for this article. Con- ers—more creative. However, companies value Ctact her at [email protected] if you experience and will pursue seasoned candidates would like to contribute to a future ComputingEdge who can demonstrate they are forward-thinking. article on computing careers. Contact Tsikerdekis Companies in the fi eld don’t take extreme risks, at [email protected]. but they also want to avoid stagnation in a globally competitive market. A candidate who has the experience, skills, and creative potential for new ideas is desirable.

ComputingEdge: What do you consider to be the best strategies for professional networking?

Tsikerdekis: Find what professionals in the fi eld are reading, fi nd out what groups they belong to, and use social media to put yourself out there for Selected CS articles and columns are also available for everyone to see. For example, maintain a current free at http://ComputingNow.computer.org. www.computer.org/computingedge 55 CAREER OPPORTUNITIES

CLOUDERA, INC. is recruiting for our support large complex campus local area custom built applications. Creation of New York, NY offi ce: Sales Engineer: networks (LAN) and wide area networks responsive web applications using mod- Work with customers, partners, & (WAN) solutions utilizing Cisco Nexus ern Javascript frameworks (AngularJS, prospects to understand & propose Switches, fi rewalls, load balancers, and JQuery, HTML5). Perform CI & automated solutions using Cloudera technologies. Infoblox DNS devices. Participate in testing using TeamCity & Robot Frame- Mail resume w/job code #37412 to: Cloud- complex network transformations and work. Build programs/batch for Adhoc era, Attn.: HR, 1001 Page Mill Rd., Bldg. 2, provide solutions. Resolve routing is- Reports via Oracle for mass data correc- Palo Alto, CA 94304. sues using the Enhanced Interior Gate- tion, analytical purposes &/or for informa- way Routing Protocol (EIGRP) and Border tional use. 40 hrs/wk. Bachelors in Comp CLOUDERA, INC. is recruiting for our Gateway Protocol (BGP). Apply to: Gerald Eng or Comp Sci or foreign equivalent & Palo Alto, CA offi ce: Product Manager: O’Mara, #82115, AHS Hospital Corp, 100 5 years exp req’d or a Masters in same. Develop creative models for pricing Madison Avenue, Morristown, NJ 07960. Send res to Brandon Associates Ltd, 26 & packaging of software solutions Sarah Drive, Farmingdale, NY 11735 using knowledge of how software is used COMPUTER PROGRAMMER – design, & competitive landscape. Mail re- create, modify, test & implement for IT Co SR. SVC CNSLT. (NY, NY & unanticip clnt sume w/job code #36162 to: Cloudera, multi-threaded client/server distributed sites thr US) Prvde proj plnng spprt. As- Attn.: HR, 1001 Page Mill Rd., Bldg. 2, Palo applications in C#, ASP.NET & C++ using sess need for cstmztn w/in a clients’ IT Alto, CA 94304. WIN 32API & Network Sockets API for Mi- envrmnt. Identfy tech implmntn issues & crosoft Windows environment based on idntfy fi xes. Engage as part of proj team DIRECTORS OF CUSTOMER SUPPORT. specs drawn up by our Stakeholders &/ & contrib prod & solution exprtse. Utilize Setup and deliver global customer sup- or Business Analyst. Develop software Java & scrptng to crte, send, receive, & port function including phone, email with Terminal Emulators for Windows us- read msgs using in-msgng techs. Collab and web portal support. Vasona Net- ing TN3270/HllAPI/VT100 – Hostbridge & interact w/ Suprt to open & track prod works Inc, San Jose, CA. hr@vasona Middleware & Qcare. Work with network issues. REQS: Bach deg or for equiv in networks.com. Travel required to unantic- protocols: TCP/IP, UDP, SIP, H.323, QOS & CS, Math, Engg (any) or a rel fi eld + 5 yrs ipated client locations internationally and RTP. Develop BizTalk apps for XSD/XSLT/ prog wrk exp in job &/or a rel occup. Will throughout US. Ref. 1A. XPATH .NET/WCF for processing Health- also accept a Master’s deg or for equiv care EDI transactions. Provide subject in CS, Math, Engg (any) or rel fi eld + 2 yrs matter expertise services & level 3 help prof wrk exp in job &/or rel occup. Must SYSTEMS ANALYST III. Design and desk support to our customers around

TECHNOLOGY SOFTWARE SOFTWARE Oracle America, Inc. Oracle America, Inc. has openings for Oracle America, Inc. has openings for SOFTWARE has openings for DEVELOPER SOFTWARE SOFTWARE positions in Austin, TX. DEVELOPER Job duties include: Design, develop, DEVELOPER troubleshoot and/or test/QA software. As positions in Morrisville, NC. positions in Morrisville, NC. a member of the software engineering division, apply knowledge of software Job duties include: Design, develop, Job duties include: Design, develop, architecture to perform tasks associated troubleshoot and/or test/QA software. troubleshoot and/or test/QA with developing, debugging, or designing Travel to various unanticipated sites software applications or operating throughout the United States required. software. systems according to provided design May telecommute from home. specifications. Apply by e-mailing resume to [email protected], Apply by e-mailing resume to Apply by e-mailing resume to [email protected], referencing 385.19726. [email protected], referencing 385.18584. referencing 385.19165. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

56 August 2016 Published by the IEEE Computer Society 2469-7087/16/$33.00 © 2016 IEEE 142 COMPUTER PUBLISHED BY THE IEEE COMPUTER SOCIETY 0018-9162/16/$33.00 © 2016 IEEE CAREER OPPORTUNITIES CAREER OPPORTUNITIES

CLOUDERA, INC. is recruiting for our support large complex campus local area custom built applications. Creation of have exp w/ Perf IT cnsltng svcs in a cli- apps, dealer systems, &large scale data cust-facing tech docs. Assist tech suprt New York, NY offi ce: Sales Engineer: networks (LAN) and wide area networks responsive web applications using mod- ent envrmnt; Archtctng, scoping, assess- processing hubs. Evaluate &install new it- in trblshtng cust issues. Archtct sols fr Work with customers, partners, & (WAN) solutions utilizing Cisco Nexus ern Javascript frameworks (AngularJS, ing & implmntng Svc Vrtliztn & App Test erations of soft &hardwr in Lead Connect clients. Collect & articulate cust feature prospects to understand & propose Switches, fi rewalls, load balancers, and JQuery, HTML5). Perform CI & automated prods; Assessing, scoping & communtng environment on Siebel Automotive. Dvlp reqsts. REQ: Bach deg or for equiv in CS, solutions using Cloudera technologies. Infoblox DNS devices. Participate in testing using TeamCity & Robot Frame- how Svc Vrtliztn & App Test solutions &improve Lead Connect Data Warehouse Math, Engg (any) or a rel field + 5 yrs prog Mail resume w/job code #37412 to: Cloud- complex network transformations and work. Build programs/batch for Adhoc resolve typical client bus needs; Sftwre using Oracle Database &OBIEE. Dvlp sys exp in job &/or rel occup. Must have exp era, Attn.: HR, 1001 Page Mill Rd., Bldg. 2, provide solutions. Resolve routing is- Reports via Oracle for mass data correc- test procss; Cstmzng solutions using C, to sys integrations. Categorize &map lead w/: Perf IT consltng svcs acr diverse cli- Palo Alto, CA 94304. sues using the Enhanced Interior Gate- tion, analytical purposes &/or for informa- C++, Java & VB.Net & scripting using VB data, usability, user interface layout &cus- ent envrmnts; XML Gateway tech; REST & way Routing Protocol (EIGRP) and Border tional use. 40 hrs/wk. Bachelors in Comp Script, Java Script & DOS; Mntrng new tomer/user exp analyzing data for optimi- SOAP web svcs; Internet & Ntwrk security CLOUDERA, INC. is recruiting for our Gateway Protocol (BGP). Apply to: Gerald Eng or Comp Sci or foreign equivalent & cnsltnts; Freq travel to unanticp client zation. Associate, Information Systems standards; Utilizing infrstrctre devices, Palo Alto, CA offi ce: Product Manager: O’Mara, #82115, AHS Hospital Corp, 100 5 years exp req’d or a Masters in same. sites in US; Wrk fr home avail anywhere or Business Administration. 3 yrs exp as identity & access mgmt sys & reliable Develop creative models for pricing Madison Avenue, Morristown, NJ 07960. Send res to Brandon Associates Ltd, 26 in US; 40 hrs/wk; Mon-fri; 8:30 am-5:30 Lead Management System Consultant or release messaging; Prog using OS-level & packaging of software solutions Sarah Drive, Farmingdale, NY 11735 pm; Send resume to: Althea Wilson, CA LMS Analyst or Lead. Exp must include scripting incl Shell scripting, Bash, XSL using knowledge of how software is used COMPUTER PROGRAMMER – design, Technologies, 201 North Franklin Street, dvlpmt of soft &technical solutions for trnsfrmtn & JavaScript, CSS & HTML; & competitive landscape. Mail re- create, modify, test & implement for IT Co SR. SVC CNSLT. (NY, NY & unanticip clnt Suite 2200, Tampa, FL, 33602, Refer to Lead Mgmt Sys soft; developing sys to Freq travel to unanticp client sites in US sume w/job code #36162 to: Cloudera, multi-threaded client/server distributed sites thr US) Prvde proj plnng spprt. As- Requisition #132384 sys integrations; categorizing &mapping & internat’ly; Wrk fr hme anywhere in the Attn.: HR, 1001 Page Mill Rd., Bldg. 2, Palo applications in C#, ASP.NET & C++ using sess need for cstmztn w/in a clients’ IT lead data, usability, user interface layout US; 40 hrs/wk; Mon-Fri; 8:30 am-5:30 pm. Alto, CA 94304. WIN 32API & Network Sockets API for Mi- envrmnt. Identfy tech implmntn issues & REQUIREMENTS ANALYST, LEAD PRO- &customer/user exp analyzing data for Send resume to: Althea Wilson, CA Tech- crosoft Windows environment based on idntfy fi xes. Engage as part of proj team CESSING & LEAD MANAGEMENT, Gen- optimization; &applying STAR &ADF data nologies, 201 North Franklin Street, Suite schemas. Mail resume to Alicia Scott- 2200, Tampa, FL, 33602, Refer to Requi- DIRECTORS OF CUSTOMER SUPPORT. specs drawn up by our Stakeholders &/ & contrib prod & solution exprtse. Utilize eral Motors Company, Detroit, MI. Lead Wears, GM Global Mobility, 300 Renais- sition #132383 Setup and deliver global customer sup- or Business Analyst. Develop software Java & scrptng to crte, send, receive, & team of 4 IT Analysts, plan &create flex- sance Center, Mail Code 482-C32-D44, port function including phone, email with Terminal Emulators for Windows us- read msgs using in-msgng techs. Collab ible Lead Management System (LMS) Detroit, MI 48265, Ref#575. and web portal support. Vasona Net- ing TN3270/HllAPI/VT100 – Hostbridge & interact w/ Suprt to open & track prod technologies, promoting ease of use SOLUTIONS ARCHITECT. Ericsson Inc. works Inc, San Jose, CA. hr@vasona Middleware & Qcare. Work with network issues. REQS: Bach deg or for equiv in interfaces between GM, dealers &cus- has an opening for the position of Solu- networks.com. Travel required to unantic- protocols: TCP/IP, UDP, SIP, H.323, QOS & CS, Math, Engg (any) or a rel fi eld + 5 yrs tomers on internet-based platforms us- SR. SVC CNSLT. (NY, NY & unanticip client tions Architect in Plano, TX to define, ipated client locations internationally and RTP. Develop BizTalk apps for XSD/XSLT/ prog wrk exp in job &/or a rel occup. Will ing &applying Standards for Technology sites thr US & internat’ly) Implmnt Layer 7 analyze & manage customer require- throughout US. Ref. 1A. XPATH .NET/WCF for processing Health- also accept a Master’s deg or for equiv in Automotive Retail (STAR) &Auto-lead prods, Mobile Access Gateway & Mobile ments utilizing Ericsson’s OSS and BSS care EDI transactions. Provide subject in CS, Math, Engg (any) or rel fi eld + 2 yrs Data Form (ADF) data schemas. Analyze Access Svcs. Dlvr on-site & remote train- portfolio. Up to 20% domestic travel re- matter expertise services & level 3 help prof wrk exp in job &/or rel occup. Must enterprise sys design, end to end data ing. Exch structured info in the implm- quired. Telecommuting is available for SYSTEMS ANALYST III. Design and desk support to our customers around streams, integration of web &mobile ntn of web svc in comp ntwrks. Produce

TECHNOLOGY SOFTWARE SOFTWARE SOFTWARE TECHNOLOGY QA ANALYST Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for has openings for has openings for SOFTWARE SYSTEMS SOFTWARE DEVELOPER SOFTWARE SOFTWARE QA positions in Austin, TX. DEVELOPMENT ADMINISTRATOR DEVELOPER Job duties include: Design, develop, DEVELOPER positions in Redwood Shores, CA. ANALYST troubleshoot and/or test/QA software. As positions in Morrisville, NC. positions in Columbia, MD. positions in . a member of the software engineering CONSULTANT Morrisville, NC Job duties include: Perform perfor- Job duties include: Develop and division, apply knowledge of software positions in Solon, OH. Job duties include: Design, develop, mance trend analysis and manage the execute software test plans. Develop, Job duties include: Design, develop, architecture to perform tasks associated troubleshoot and/or test/QA software. Job duties include: Analyze requirements server/network capacity. Monitor with developing, debugging, or designing apply and maintain quality standards troubleshoot and/or test/QA Travel to various unanticipated sites and deliver functional and technical production’s environmental health and software applications or operating for company products with adher- throughout the United States required. solutions. Implement products and software. systems according to provided design resource usage; identify potential ence to both internal and external May telecommute from home. technologies to meet post-sale customer specifications. issues. standards. Apply by e-mailing resume to needs. Apply by e-mailing resume to [email protected], Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], [email protected], referencing 385.19726. [email protected], [email protected], referencing 385.18584. referencing 385.19165. referencing 385.18241. referencing 385.19529. referencing 385.18275. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

this position from anywhere in the United equiv+5yr progressive exp. Travel/relo in Oracle Application Financials and SCM States. To apply please mail resume to req. Send resume & applied position to: domain in P2P (Procure to pay) and O2C Ericsson Inc. 6300 Legacy Drive, R1-C12 [email protected] or 460 Park (Order to Cash) modules GL, AP, AR, TCA, Plano, TX 75024 and indicate applying Ave. S., Ste# 1101, New York, NY 10016 PO, OM, INV and SLA. Send resume: In- for 16-TX-3091. Attn: Recruit. dotronix Int.l Corp., Recruiting (MS), 331 Main St, Poughkeepsie, NY 12601. MPHASIS CORP has multi openings at ORACLE APPLICATIONS EBS LEAD F/T various levels for the follow’g positions (Poughkeepsie, NY) Position involves SPLUNK INC. has the following job op- at its office in NY, NY & unanticipated cli- travel to various unanticipated worksites portunities in San Francisco, CA: Soft- ent sites thr/o the US 1. Info. Sys. Anyst* up to 100% of the time anywhere in the ware Engineer (Req#9TL4GC) Create - Ana. & provide sys req & spec. 2. SW United States. Must have Bach deg or the fault-tolerant distributed systems in Dvlper* - Design, dvlp & modify SW sys. foreign equiv in Electronic Engg, Technol- multi-threaded & multi-processed en- 3. Sys. Architect Dvlper* - Dvlp IT archi - ogy, or related w/5 yrs of progressive exp vir. Software Engineer (Req#9U23ZS) tecture 4. Graphic UI Desgr* - Design UI managing and Leading a team of 3 devel- Perform auto testing of solutions & apps & perform UAT 5. N/W Infra Eng* - Main- opers in Oracle Application R12 analysis, dev for IT service monitoring & ops intel tain & TRBL n/w, design, dvlp, install design, development, implementation, on prod platform. Senior Software Engi- n/w infra appl. 6. Business Operation Upgrade, testing and performance tun- neer (Req#9W9M3F) Architect, design, Anyst* - Ana bus process thru app of s/w ing through full product development life & dev sw for provisioning, mngmnt, & sol. 7. IT Mgr* - Plan & manage the de - cycle and release process. Mentor the ju - monitoring of distributed sw sys. Splunk livery of IT proj. 8. Enterprise Svc Enga- nior team members by providing technol- Inc. has the following job opportunity in gem’t Mgr* - E2E sale of IT svc/prod. 9. ogy knowledge transition. Provide com- Cupertino, CA: Senior Analyst, SOX IT Eng Engagem’t Mgr* - Manage & direct prehensive consultation to business unit & Internal Audit (Req#9XXPF9) Execute business integration of proj activities. 10. & IT management and staff at the highest IT Audit SOX prgrm & rev sys implemen- Mkt Dvlpt Mgr* - Promote IT svc/prod. technical level on all phases of applica- tations, apps, & IT-rel risk areas. Refer & impl bus plans. Must have a Bachelor/ tion. Provide subject matter expertise to Req# & mail resume to Splunk Inc., equiv and prior rel. exp, Master/equiv, or and implement the code using following ATTN: J. Aldax, 250 Brannan Street, San Master/equiv and prior rel. exp. Edu/exp tools/technologies: SQL, PLSQL, Re- Francisco CA 94107. Individuals seeking req vary depending on position level/ ports, Discover, Interfaces, Conversions, employment at Splunk are considered type. *Lead positions in this occupation Data Migrations, Forms, Toad, Workflow, without regards to race, religion, color, must have Master/equiv+2yr or Bach/ XML Reports, SQL Loader and Oracle 10g national origin, ancestry, sex, gender,

TECHNICAL SOFTWARE TECHNICAL Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for TECHNICAL TECHNICAL ANALYST - SUPPORT SOFTWARE ANALYST positions in Orlando, FL. DEVELOPER – SUPPORT Job duties include: Deliver post-sales positions in Colorado Springs, CO. support and solutions to the Oracle positions in Nashua, NH. customer base while serving as an Job duties include: Deliver post-sales support advocate for customer needs. Research, Job duties include: Design, develop, and solutions to the Oracle customer base resolve, and respond to complex customer troubleshoot and/or test/QA while serving as an advocate for customer inquiries and technical questions regard- software. needs. Research, resolve, and respond to ing the use of, and troubleshooting for, complex customer inquiries and technical questions regarding the use of, and trouble- our product line. Apply by e-mailing resume to shooting for, our product line. Apply by e-mailing resume to [email protected], referencing 385.16166. Apply by e-mailing resume to [email protected], [email protected], referencing 385.19658. Oracle supports workforce diversity. referencing 385.19592. Oracle supports workforce diversity. Oracle supports workforce diversity.

58 ComputingEdge August 2016 144 COMPUTER WWW.COMPUTER.ORG/COMPUTER CAREER OPPORTUNITIES CAREER OPPORTUNITIES

this position from anywhere in the United equiv+5yr progressive exp. Travel/relo in Oracle Application Financials and SCM gender identity, gender expression, sex- procurement, strategic sourcing, good job offered. Mail resume to Alicia Scott- States. To apply please mail resume to req. Send resume & applied position to: domain in P2P (Procure to pay) and O2C ual orientation, marital status, age, phys- receipt/inbound deliveries &business Wears, Global Mobility, 300 Renaissance Ericsson Inc. 6300 Legacy Drive, R1-C12 [email protected] or 460 Park (Order to Cash) modules GL, AP, AR, TCA, ical or mental disability or medical con - intelligence reporting. Work on global Center, MC:482-C32-D44, Detroit, MI Plano, TX 75024 and indicate applying Ave. S., Ste# 1101, New York, NY 10016 PO, OM, INV and SLA. Send resume: In- dition (except where physical fitness is a implementation for deployment of SAP 48265, Ref#3482. for 16-TX-3091. Attn: Recruit. dotronix Int.l Corp., Recruiting (MS), 331 valid occupational qualification), genetic solution. Mentoring teams responsible Main St, Poughkeepsie, NY 12601. information, veteran status, or any other for blueprint, configuration &testing. Re- CLOUDERA, INC. is recruiting for our MPHASIS CORP has multi openings at ORACLE APPLICATIONS EBS LEAD F/T consideration made unlawful by federal, sponsible for analyzing, designing, dvlpg, New York, NY office: Professional Ser- various levels for the follow’g positions (Poughkeepsie, NY) Position involves SPLUNK INC. has the following job op- state or local laws. To review US DOL’s testing &supporting post go-live for en- vices Practice Director: Manage sales & at its office in NY, NY & unanticipated cli- travel to various unanticipated worksites portunities in San Francisco, CA: Soft- EEO is The Law notice please visit: https:// tire procurement &inventory mgmt cycle delivery of professional services (consult- ent sites thr/o the US 1. Info. Sys. Anyst* up to 100% of the time anywhere in the ware Engineer (Req#9TL4GC) Create careers.jobvite.com/Splunk/EEO_poster. which includes procurement, inbound ing) to Cloudera customers for a defined - Ana. & provide sys req & spec. 2. SW United States. Must have Bach deg or the fault-tolerant distributed systems in pdf. To review Splunk’s EEO Policy State- &outbound deliveries, invoicing, ERS regional territory. Recruit, retain, manage Dvlper* - Design, dvlp & modify SW sys. foreign equiv in Electronic Engg, Technol- multi-threaded & multi-processed en- ment please visit: http://careers.jobvite. process &goods movement. Master, IT/ and mentor a team of technical consul- 3. Sys. Architect Dvlper* - Dvlp IT archi - ogy, or related w/5 yrs of progressive exp vir. Software Engineer (Req#9U23ZS) com/Careers/Splunk/EEO-Policy-State- IS or Computer Science. 36 mos exp as tants. Mail resume w/job code #34478 to: tecture 4. Graphic UI Desgr* - Design UI managing and Leading a team of 3 devel- Perform auto testing of solutions & apps ment.pdf. Pursuant to the San Francisco Business Analyst, SAP Analyst, SAP Con- Cloudera, Attn.: HR, 1001 Page Mill Rd., & perform UAT 5. N/W Infra Eng* - Main- opers in Oracle Application R12 analysis, dev for IT service monitoring & ops intel Fair Chance Ordinance, we will consider sultant working on implementation &sup- Bldg. 2, Palo Alto, CA 94304. tain & TRBL n/w, design, dvlp, install design, development, implementation, on prod platform. Senior Software Engi- for employment qualified applicants with port of SAP modules to enhance business arrest and conviction records. functionality, performance, design, con- n/w infra appl. 6. Business Operation Upgrade, testing and performance tun- neer (Req#9W9M3F) Architect, design, JUNIPER NETWORKS is recruiting for figuration &testing in SAP Modules such Anyst* - Ana bus process thru app of s/w ing through full product development life & dev sw for provisioning, mngmnt, & our Sunnyvale, CA office: Software En- as SRM, SAP MM, Business Reporting, &/ sol. 7. IT Mgr* - Plan & manage the de - cycle and release process. Mentor the ju - monitoring of distributed sw sys. Splunk SENIOR SAP ANALYST, ROSWELL, GA. gineer Staff: Design and implement cus- or SAP Sourcing &Catalog Content Mgmt livery of IT proj. 8. Enterprise Svc Enga- nior team members by providing technol- Inc. has the following job opportunity in General Motors. Work on different SAP tomer use cases for Juniper Open Con- modules. Will accept Bachelor or foreign gem’t Mgr* - E2E sale of IT svc/prod. 9. ogy knowledge transition. Provide com- Cupertino, CA: Senior Analyst, SOX IT modules to support Supply Chain &Pro - trail platform for enterprise & service equivalent, in IT/IS or CS, followed by at Eng Engagem’t Mgr* - Manage & direct prehensive consultation to business unit & Internal Audit (Req#9XXPF9) Execute curement process. Design &dvlp solution provider customers. Mail single-sided least 5 yrs progressive exp in specialty, business integration of proj activities. 10. & IT management and staff at the highest IT Audit SOX prgrm & rev sys implemen- for complete Request to Pay &Inventory resume with job code #6584 to: Juniper in lieu of required education &exp. Will Mkt Dvlpt Mgr* - Promote IT svc/prod. technical level on all phases of applica- tations, apps, & IT-rel risk areas. Refer Mgmt cycle using SAP modules MM, SRM, Networks, Attn.: MS A.8.429A, 1133 Inno- also accept any equally suitable combi- & impl bus plans. Must have a Bachelor/ tion. Provide subject matter expertise to Req# & mail resume to Splunk Inc., SAP Sourcing, WM, &SD. Design, configu- vation Way, Sunnyvale, CA 94089. equiv and prior rel. exp, Master/equiv, or and implement the code using following ATTN: J. Aldax, 250 Brannan Street, San ration &testing in SAP Modules SRM, MM nation of education, training, &/or exp Master/equiv and prior rel. exp. Edu/exp tools/technologies: SQL, PLSQL, Re- Francisco CA 94107. Individuals seeking &Business Reporting for indirect material which would qualify applicant to perform req vary depending on position level/ ports, Discover, Interfaces, Conversions, employment at Splunk are considered type. *Lead positions in this occupation Data Migrations, Forms, Toad, Workflow, without regards to race, religion, color, must have Master/equiv+2yr or Bach/ XML Reports, SQL Loader and Oracle 10g national origin, ancestry, sex, gender, Cisco Systems, Inc. is accepting resumes for the following positions:

TECHNICAL COLUMBIA, MD: Software Engineer (Ref.# COLU1): SAN FRANCISCO, CA: Manager, Engineering (Ref.# SOFTWARE TECHNICAL Responsible for the definition, design, development, test, SF26): Schedule and evaluate the resources required for debugging, release, enhancement or maintenance of networking multiple projects in terms of human resources and hardware Oracle America, Inc. Oracle America, Inc. software. Technical Leader (Ref.# COLU8): Lead equipment allocation. Oracle America, Inc. has openings for engineering groups on projects to design, develop or test has openings for hardware or software products. SAN JOSE/MILPITAS/SANTA CLARA, CA: Manager, has openings for Engineering (Ref.# SJ4): Schedule and evaluate the DENVER, CO: Software/QA Engineer (Ref.# DEN2): resources required for multiple projects in terms of human TECHNICAL TECHNICAL Debug software products through the use of systematic tests resources and hardware equipment allocation. Technical to develop, apply, and maintain quality standards for company Marketing Engineer (Ref.# SJ178): Responsible ANALYST - SUPPORT SOFTWARE products. for enlarging company’s market and increasing revenue by ANALYST marketing, supporting, and promoting company’s technology positions in Orlando, FL. MIAMI, FL: Systems Engineer (Ref.# MIA2): Provide to customers. Travel may be required to various unanticipated – business-level guidance to the account team or operation on locations throughout the United States. Virtual Systems DEVELOPER SUPPORT technology trends and competitive threats, both at a technical Engineer (Ref.# SJ544): Engage customers virtually Job duties include: Deliver post-sales and business level. via collaborative tools and technology including web and positions in Colorado Springs, CO. video conferencing. IT Manager (Ref.# SJ182): Drive support and solutions to the Oracle positions in Nashua, NH. improvements to the strategic, core and support processes RESEARCH TRIANGLE PARK, NC: Solutions of Company’s Commerce using data and analytics. User Job duties include: Deliver post-sales support Consultant (Ref.# RTP815): Perform as a technical customer base while serving as an Experience Designer/User Experience Engineer/ lead to design, migrate, deploy and support deployments on Job duties include: Design, develop, and solutions to the Oracle customer base User Centered Design Engineer (Ref.# SJ587): advocate for customer needs. Research, company platforms. Telecommuting permitted and travel may Identify user interaction requirements and develop user while serving as an advocate for customer be required to various unanticipated locations throughout the resolve, and respond to complex customer troubleshoot and/or test/QA experience interface specifications and guidelines. Software United States. Customer Support Engineer (Ref.# needs. Research, resolve, and respond to Engineer (Ref.# SJ10): Responsible for the definition, inquiries and technical questions regard- Responsible for providing technical support regarding software. RTP1): design, development, test, debugging, release, enhancement or complex customer inquiries and technical the company’s proprietary systems and software. ing the use of, and troubleshooting for, maintenance of networking software. questions regarding the use of, and trouble- our product line. ROSEMONT, IL: Consulting Systems Engineer Apply by e-mailing resume to PLEASE MAIL RESUMES WITH REFERENCE NUMBER shooting for, our product line. (Ref.# ROSE17): Provide specific end-to-end solutions and TO CISCO SYSTEMS, INC., ATTN: M51H, 170 W. Tasman [email protected], architecture consulting, technical and sales support for major Apply by e-mailing resume to Drive, Mail Stop: SJC 5/1/4, San Jose, CA 95134. No phone calls account opportunities at the theater, area, or operation level. Apply by e-mailing resume to please. Must be legally authorized to work in the U.S. without [email protected], referencing 385.16166. Travel may be required to various unanticipated locations [email protected], sponsorship. EOE. referencing 385.19658. throughout the United States. Oracle supports workforce diversity. referencing 385.19592. www.cisco.com Oracle supports workforce diversity. Oracle supports workforce diversity.

www.computer.org/computingedge 59 144 COMPUTER WWW.COMPUTER.ORG/COMPUTER AUGUST 2016 145 ERICSSON INC. has multiple openings in Bellevue, WA for the following positions: TEST ENGINEER _ to support customer test case design, development, and test case execution activities. Job ID 16-WA-3638. SOLUTIONS ARCHI- TECT _ to develop solution descriptions & statements of work & solution cost estimates. Job ID 16-WA-3727. ENGI- NEER _ to perform system verification Juniper Networks is recruiting & testing of patches & configuration. for our Sunnyvale, CA office: Telecommuting is available for this po - sition from anywhere in the U.S. Job ID 16-WA-2238. TEST ENGINEER _ to im - Technical Support Engineering tomer requirements to software plement testing process, develop and Staff #18103: Be the focal tech- design. execute testing and verification. Up to 15% domestic and international travel. nical support contact & handle ASIC Engineer #34265: Define, Job ID 16-WA-3077. To apply please mail resume to Ericsson Inc. 6300 Leg- high-priority issues for Juniper’s architect, code, and deliver ver- Advanced Service customers acy Drive, R1-C12 Plano, TX 75024 and ification suites/tests for ASICs indicate Job ID of interest. to ensure prompt restoration & that enable faster, denser, and resolution. feature-rich systems. Write SOFTWARE Sales Demonstration Engineer complete verification plans in- Staff #22850: Develop and deliv- dependently. Oracle America, Inc. er detailed support of Proof of Technical Support Engineer has openings for Concept (POC) testing, based #17355: Provide technical assis- upon the test plans received tance & conduct troubleshoot- from company customers and ing to resolve customer hard- SOFTWARE prospects. ware & software issues. Identify Technical Support Engineer service support initiatives to im- DEVELOPMENT #20295: Become the focal tech- prove the supportability of pro- nical support contact & han- cesses, products & systems. CONSULTANT dle high-priority issues for Ad- Software Engineer Sr. Staff vanced Service customers on #10828: Design, develop, trou- positions in New York, NY. Juniper products. Take owner- bleshoot and debug software ship on high-priority customer programs for enhancements Job duties include: Analyze require- issues & ensure prompt service and new products. Architect ments and deliver functional and restoration and resolution. and develop complex packet technical solutions. Software Engineer Staff #8566: forward features, such as IPv4/ Design, develop, and maintain IPv6/MPLS, Multicast, Routing, Apply by e-mailing resume to packet forwarding code on Ju- Firewall/Filter, Cos, and Service niper’s products. Translate cus- area. [email protected], Juniper Networks is recruiting referencing 385.18240. for our Westford, MA office: Oracle supports workforce diversity. ASIC Engineer #30172: Develop es for cutting edge technology detailed test plans and block networking chips using System and system-level test bench- Verilog and C++. CLASSIFIED LINE AD SUBMISSION DETAILS: Mail single-sided resume with job code # to Rates are $425.00 per column inch ($640 minimum). Eight lines per column inch Juniper Networks and average five typeset words per line. Attn: MS A.8.429A Send copy at least one month prior to publication date to: 1133 Innovation Way DEBBIE SIMS Sunnyvale, CA 94089 Classified Advertising Email: [email protected]

60 ComputingEdge August 2016 146 COMPUTER WWW.COMPUTER.ORG/COMPUTER ERICSSON INC. has multiple openings in Bellevue, WA for the following posi- It’s work that matters. It's what we do at Symantec. Symantec is the world leader in providing solutions to help individuals and enterprises assure the security, availability, and tions: TEST ENGINEER _ to support cus- integrity of their information. In essence, we protect the free flow of information in a connected world. As the fourth largest independent software company in the world, tomer test case design, development, Symantec has operations in more than 40 countries with 475 out of Fortune's global 500 companies using our solutions. People look to us to safeguard the integrity of their informa- and test case execution activities. Job ID 16-WA-3638. SOLUTIONS ARCHI- tion, ensuring it is secure and available. Achieving this ambitious goal is only possible through the combined efforts of the innovators and visionaries that Symantec continuously TECT _ to develop solution descriptions attracts. Symantec draws the very best people with a variety of backgrounds, experiences and perspectives and provides them with a work environment where uniqueness is & statements of work & solution cost valued and empowered. The creative people we attract help define the spirit of innovation at Symantec. Symantec is proud to be an equal opportunity employer. estimates. Job ID 16-WA-3727. ENGI- NEER _ to perform system verification Juniper Networks is recruiting & testing of patches & configuration. Symantec Corporation currently has openings for the following positions in Mountain View, CA (various levels/types): for our Sunnyvale, CA office: Telecommuting is available for this po - sition from anywhere in the U.S. Job ID 16-WA-2238. TEST ENGINEER _ to im - Computer Systems Analysts (CSAHQ716) Analyze engineering, business and/or other business intelligence issues for application to Symantec solutions; and provide operational Technical Support Engineering tomer requirements to software plement testing process, develop and support in the development and implementation process of computer software applications, systems or services. Staff #18103: Be the focal tech- design. execute testing and verification. Up to 15% domestic and international travel. Database Engineers (DBEHQ716) Support all non-product and product databases including installation, configuration, upgrade, backup and recovery. Must be available to work nical support contact & handle ASIC Engineer #34265: Define, Job ID 16-WA-3077. To apply please mail resume to Ericsson Inc. 6300 Leg- on projects at various, unanticipated sites throughout the United States. May telecommute from home. high-priority issues for Juniper’s architect, code, and deliver ver- Advanced Service customers acy Drive, R1-C12 Plano, TX 75024 and ification suites/tests for ASICs indicate Job ID of interest. Engineering Managers (EMHQ716) Direct and supervise team of engineering (QA and/or development teams). Develop standards for products and/or oversee development and to ensure prompt restoration & that enable faster, denser, and execution of software and/or analysis of test results. resolution. feature-rich systems. Write SOFTWARE IT Architects (ITAHQ716) Technically guide the customer engineering team on designing, developing, deploying and operating their solutions and services by offering OpenStack Sales Demonstration Engineer complete verification plans in- cloud expertise. Staff #22850: Develop and deliv- dependently. Oracle America, Inc. er detailed support of Proof of Technical Support Engineer has openings for IT Business Analysts (ITBAHQ716) Participate in creation and definition of platform roadmaps. Concept (POC) testing, based #17355: Provide technical assis- upon the test plans received Product Managers (PDMHQ716) Participate in all software product development life cycle activities. Move software products through the product development cycle from tance & conduct troubleshoot- design and development to implementation and testing. from company customers and ing to resolve customer hard- SOFTWARE prospects. ware & software issues. Identify Program Managers (PMHQ716) Work closely with engineering members, managers, and leads, product managers, ensure rapid execution and on time, high quality delivery of Technical Support Engineer service support initiatives to im- DEVELOPMENT complex Data Loss Prevention (DLP) projects. #20295: Become the focal tech- prove the supportability of pro- Service Managers- IT Hosting (SMHQ716) Responsible for IT Hosting technologies that may include but not be limited to Infrastructure as a service, Compute, Storage and nical support contact & han- cesses, products & systems. CONSULTANT backup, Enablement services, Infrastructure Middleware). dle high-priority issues for Ad- Software Engineer Sr. Staff vanced Service customers on #10828: Design, develop, trou- positions in New York, NY. Software Engineers (SWEHQ716) Responsible for analyzing, designing, debugging and/or modifying software; or evaluating, developing, modifying, and coding software programs to support programming needs. Juniper products. Take owner- bleshoot and debug software ship on high-priority customer programs for enhancements Job duties include: Analyze require- issues & ensure prompt service Software QA Engineers (SQAHQ716) Responsible for developing, applying and maintaining quality standards for company products. Develop and execute software test plans. and new products. Architect ments and deliver functional and Analyze and write test standards and procedures. restoration and resolution. and develop complex packet technical solutions. Software Engineer Staff #8566: forward features, such as IPv4/ Symantec Corporation currently has openings for the following positions in San Francisco, CA (various levels/types): Design, develop, and maintain IPv6/MPLS, Multicast, Routing, Apply by e-mailing resume to Software Engineers (SWESF716) Responsible for analyzing, designing, debugging and/or modifying software; or evaluating, developing, modifying, and coding software programs packet forwarding code on Ju- Firewall/Filter, Cos, and Service to support programming needs. niper’s products. Translate cus- area. [email protected], Juniper Networks is recruiting referencing 385.18240. Symantec Corporation currently has openings for the following positions in Springfield, OR (various levels/types): for our Westford, MA office: Oracle supports workforce diversity. Software Engineers (SWEOR716) Responsible for analyzing, designing, debugging and/or modifying software; or evaluating, developing, modifying, and coding software ASIC Engineer #30172: Develop es for cutting edge technology programs to support programming needs. detailed test plans and block networking chips using System and system-level test bench- Verilog and C++. Submit resume to [email protected] . Must reference position & code listed above. EOE. For additional information about Symantec and other positions visit our website at http://www.symantec.com. CLASSIFIED LINE AD SUBMISSION DETAILS: Mail single-sided resume with job code # to Rates are $425.00 per column inch ($640 minimum). Eight lines per column inch Juniper Networks and average five typeset words per line. Attn: MS A.8.429A Send copy at least one month prior to publication date to: 1133 Innovation Way DEBBIE SIMS Sunnyvale, CA 94089 Classified Advertising Email: [email protected]

www.computer.org/computingedge AUGUST 2016 61147 146 COMPUTER WWW.COMPUTER.ORG/COMPUTER TECHNOLOGY TECHNICAL TECHNICAL Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for TECHNICAL PROGRAMMER ANALYST TECHNICAL ANALYST positions in Orlando, FL. ANALYST positions in Nashua, NH. Job duties include: Analyze user require- positions in Lehi, UT. ments to develop, implement, and/or Job duties include: Analyze user support Oracle’s global infrastructure. As a Job duties include: Analyze user requirements to develop, imple- member of the IT organization, assist with requirements to develop, implement, ment, and/or support Oracle’s global the design, development, modifications, and/or support Oracle’s global infrastructure. debugging, and evaluation of programs infrastructure. for use in internal systems within a Apply by e-mailing resume to specific function area.. Apply by e-mailing resume to [email protected], [email protected], Apply by e-mailing resume to referencing 385.18375. [email protected], referencing 385.19553. referencing 385.19280. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

TECHNICAL SOFTWARE QA ANALYST Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for TECHNICAL SOFTWARE QA ANALYST ANALYST DEVELOPER -PRODEV - SUPPORT positions in Atlanta, GA. positions in Reston, VA. positions in Lehi, UT. Job duties include: Design Job duties include: Design, develop, functional, integration, and regres- Job duties include: Deliver post-sales troubleshoot and/or test/QA software support and solutions to the Oracle sion test plans. Build and execute programs for enhancements and new manual and automated tests. customer base while serving as an products. advocate for customer needs. Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], [email protected], referencing 385.17694. referencing 385.18317. referencing 385.19690. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

62 ComputingEdge August 2016 148 COMPUTER WWW.COMPUTER.ORG/COMPUTER TECHNOLOGY TECHNICAL TECHNICAL TECHNOLOGY TECHNICAL SOFTWARE Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for has openings for has openings for has openings for TECHNICAL USER TECHNICAL PROGRAMMER ANALYST TECHNICAL EXPERIENCE ANALYST SOFTWARE ANALYST positions in Orlando, FL. ANALYST DEVELOPER DEVELOPER positions in Redwood Shores, CA. - SUPPORT positions in Lehi, UT. positions in Nashua, NH. Job duties include: Analyze user require- positions in Orlando, FL. positions in Naples, FL. ments to develop, implement, and/or Job duties include: Develop and apply Job duties include: Analyze user software design/usability processes in the support Oracle’s global infrastructure. As a Job duties include: Analyze user Job duties include: As a member of the Job duties include: Design, develop, investigation of technical problems. May requirements to develop, imple- member of the IT organization, assist with requirements to develop, implement, Support organization, deliver post-sales troubleshoot and/or test/QA develop user profiles, with emphasis on ment, and/or support Oracle’s global the design, development, modifications, and/or support Oracle’s global support and solutions to the Oracle software. human error control, display issues, visual infrastructure. debugging, and evaluation of programs infrastructure. customer base while serving as an interaction, physical manipulation and for use in internal systems within a advocate for customer needs. Apply by e-mailing resume to specific function area.. task and objective analyses. Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], Apply by e-mailing resume to Apply by e-mailing resume to [email protected], Apply by e-mailing resume to referencing 385.18258. [email protected], [email protected], referencing 385.18375. [email protected], referencing 385.19553. referencing 385.18528. referencing 385.19601. referencing 385.19280. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

TECHNICAL SOFTWARE QA ANALYST TECHNOLOGY Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. LinkedIn Corp. has openings in our Mtn View, CA location for: has openings for has openings for has openings for Software Engineer (All Levels/Types) (SWE616MV) Design, develop & integrate cutting-edge software technologies; Senior TECHNICAL QA ANALYST Engineer, Data Services Operations (6597.1170) Serve as a primary point responsible the overall health, performance and capacity of SOFTWARE back-end Hadoop based data warehouse environment; Manager, Software Engineer (6597.304) Manage the performance & career ANALYST -PRODEV development of a small team of engineers, & own significant parts of the LinkedIn anti-abuse systems that require design, architecture, DEVELOPER & coding; Senior User Interface Engineer (6597.1579) Build rich, dynamic client-side interfaces using JavaScript, while leveraging - SUPPORT positions in Atlanta, GA. positions in Reston, VA. technologies like HTML5, CSS3, Scala, Dust & Play. positions in Lehi, UT. Job duties include: Design Job duties include: Design, develop, functional, integration, and regres- LinkedIn Corp. has openings in our Sunnyvale, CA location for: Job duties include: Deliver post-sales troubleshoot and/or test/QA software sion test plans. Build and execute Software Engineer (All Levels/Types) (SWE616SV) Design, develop & integrate cutting-edge software technologies; support and solutions to the Oracle Salesforce.com Developer (6597.958) Develop, enhance, debugs, support, analyze, maintain & test new/existing functionality which programs for enhancements and new manual and automated tests. customer base while serving as an products. supports internal business units or supporting functions; Applied Research Engineer (6597.989) Design, develop & integrate cutting- advocate for customer needs. edge software technologies. Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], LinkedIn Corp. has openings in our New York, NY location for: [email protected], referencing 385.17694. referencing 385.18317. Senior Software Engineer (6597.711) Design, develop, and integrate cutting-edge software technologies. referencing 385.19690. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Please email resume to: [email protected]. Must ref. job code above when applying.

www.computer.org/computingedge AUGUST 2016 14963 148 COMPUTER WWW.COMPUTER.ORG/COMPUTER TECHNOLOGY Intuit Inc. has openings for the following positions in Santa Clara County, including Mountain View, California or any office within normal commuting distance:

Staff Systems Engineers (Job code: I-463): Apply a full understanding of software engineering methodologies and industry best practices. Utilize specialized knowledge to design, develop, and maintain Intuit’s software. Senior Appli- cation Operations Engineers (Job code: I-286): Exercises judgment in selecting methods and techniques to deliver operational solutions. Senior Software Engineers (Job code: I-190): Exercise senior level knowledge in selecting methods and techniques to design, implement, modify and support a variety of software products and services to meet user or system specifications.

Positions located in San Diego, California: Staff Network Engineers (Job code: I-2304): Work with our business units to solve for performance and security needs in the environment, and help them fix network and non-network related issues. Senior Software Engineers in Quality (Job code: I-392): Apply senior level software engineering practices and procedures to design, influence, and drive quality and testability of products and services. Staff Software Engineers (Job code: I-1968): Apply master level software engineering and industry best practices to design, implement, and support software products and services. Senior Application Operations Engineers (Job code: I-332): Exercises judgment in selecting methods and techniques to deliver operational solutions. Engages the broader team to understand and apply best practices in the selection process.

Positions located in Woodland Hills, California: Lead Financial Analysts (Job code: I-1067): Writing SQL queries to pull pricing and payments data to conduct trends analysis across various payments product offerings. Senior Software Engineers (Job code: I-484): Exercise senior level knowledge in selecting methods and techniques to design, implement, modify and support a variety of software products and services to meet user or system specifications.

Positions located in Plano, Texas: Staff Systems Engineers (Job code: I-247): Support and Engineer enterprise wide Server Automation platform. Senior Software Engineers (Job code: I-34): Exercise senior level knowledge in selecting methods and techniques to design, implement, modify and support a variety of software products and services to meet user or system specifications. Staff Software Engineers (Job code: I-930): Apply master level software engineering and industry best practices to design, implement, and support software products and services.

To apply, submit resume to Intuit Inc., Attn: Olivia Sawyer, J203-6, 2800 E. Commerce Center Place, Tucson, AZ 85706. You must include the job code on your resume/cover letter. Intuit supports workforce diversity.

64150 COMPUTER ComputingEdge WWW.COMPUTER.ORG/COMPUTERAugust 2016 TECHNOLOGY TECHNOLOGY SOFTWARE SOFTWARE Intuit Inc. Oracle America, Inc. Oracle America, Inc. Oracle America, Inc. has openings for has openings for has openings for the following positions in Santa Clara County, including has openings for Mountain View, California or any office within normal commuting distance: SOFTWARE MANAGING DEVELOPMENT SOFTWARE Staff Systems Engineers (Job code: I-463): Apply a full understanding of software engineering methodologies and PRINCIPAL CONSULTANT industry best practices. Utilize specialized knowledge to design, develop, and maintain Intuit’s software. Senior Appli- positions in Ann Arbor, MI. DEVELOPER positions in West Conshohocken, PA. cation Operations Engineers (Job code: I-286): Exercises judgment in selecting methods and techniques to deliver CONSULTANT Job duties include: Design, develop, operational solutions. Senior Software Engineers (Job code: I-190): Exercise senior level knowledge in selecting positions in Solon, OH. troubleshoot and debug software programs for databases, applications, Job duties include: Design, develop, methods and techniques to design, implement, modify and support a variety of software products and services to Job duties include: Analyze requirements tools, networks etc. You will assist in troubleshoot and/or test/QA meet user or system specifications. and deliver functional and technical defining and developing software for software. May telecommute from solutions. Implement products and tasks associated with the developing, home. Positions located in technologies to meet post-sale customer San Diego, California: debugging or designing of software needs. Staff Network Engineers (Job code: I-2304): Work with our business units to solve for performance and security applications or operating systems. Apply by e-mailing resume to needs in the environment, and help them fix network and non-network related issues. Senior Software Engineers in Apply by e-mailing resume to [email protected], Apply by e-mailing resume to Quality (Job code: I-392): Apply senior level software engineering practices and procedures to design, influence, and [email protected], referencing 385.19391. referencing 385.18247. [email protected], drive quality and testability of products and services. Staff Software Engineers (Job code: I-1968): Apply master referencing 385.18238. Oracle supports workforce diversity. level software engineering and industry best practices to design, implement, and support software products and Oracle supports workforce diversity. Oracle supports workforce diversity. services. Senior Application Operations Engineers (Job code: I-332): Exercises judgment in selecting methods and techniques to deliver operational solutions. Engages the broader team to understand and apply best practices in the Western Digital Technologies, Inc. selection process. has openings for the following positions in Milpitas, California:

Positions located in Woodland Hills, California: Design Engineers (Job code: WD439): Design, develop, modify and evaluate electronic parts, components or integrated circuitry for electronic equipment and other hardware systems. Firmware Engineers (Job code: WD942): Analyze, design, program, debug, troubleshoot and/or modify software for firmware (IC embedded code) Lead Financial Analysts (Job code: I-1067): Writing SQL queries to pull pricing and payments data to conduct trends applications. Industrial & Operations Engineers (Job code: WD634): Responsible for the development of estimated or engineering standards, optimum process analysis across various payments product offerings. Senior Software Engineers (Job code: I-484): Exercise senior layouts and methods. Sr. Test Development Engineers (Job code: WD186): Responsible for designing, developing, and implementing cost-effective methods of testing and troubleshooting systems and equipment. Device Engineers (Job code: WD609): Design, develop, modify and evaluate digital electronic parts, components or level knowledge in selecting methods and techniques to design, implement, modify and support a variety of software integrated circuitry for digital electronic equipment and other hardware systems. Sr. Research Engineers (Job code: WD369): Responsible for validating LDPC functions products and services to meet user or system specifications. in controller ASIC that will be used for all new ESS products introduced in 2016. Principal Process Integration Engineers (Job code:WD604): Responsible for full device process integration including creating, maintaining and developing current and future process technologies. Must be available to work on projects at various, unanticipated sites throughout the United States. Sr. Logic Design Engineers (Job code: WD978): Designing, optimizing, verifying and debugging complex logic blocks for Positions located in Plano, Texas: emerging 3D non-volatile memory. Programmers/Analysts (Job code: WD504): Analyze complex business and business process problems and come up with solutions Staff Systems Engineers (Job code: I-247): Support and Engineer enterprise wide Server Automation platform. which can be either manual or with automated system or a combination. Product Engineers (Job code: WD312): Provide production support engineering for a specific Senior Software Engineers (Job code: I-34): Exercise senior level knowledge in selecting methods and techniques product or group of products after transfer from design to high volume production. Product Engineers (Job code: WD253): Provide production support engineering for a specific product or group of products after transfer from design to high volume production. Some travel may be required to work on projects at various, unanticipated to design, implement, modify and support a variety of software products and services to meet user or system specifi- sites throughout the United States. cations. Staff Software Engineers (Job code: I-930): Apply master level software engineering and industry best Positions in Longmont, Colorado: practices to design, implement, and support software products and services. Staff ASIC Verification Engineers (Job code: WD912): Responsible for building module level design verification environment and verifying ASIC design.

To apply, reference job code # & mail resume to: To apply, submit resume to Intuit Inc., Attn: Olivia Sawyer, J203-6, 2800 E. Commerce Center Place, Tucson, AZ 85706. Western Digital Technologies, Inc., 951 SanDisk Drive, MS: HRGM, Milpitas, CA 95035. You must include the job code on your resume/cover letter. Intuit supports workforce diversity. Must be legally auth to work in the U.S. w/out sponsorship. EOE

150 COMPUTER WWW.COMPUTER.ORG/COMPUTER www.computer.org/computingedge 65 AUGUST 2016 151 CAREER OPPORTUNITIES

APPLE INC. has the following job opportunities in Cupertino, CA:

ENGINEERING (Req#9FZNSB) Evaluate iPhone tomate app & infrastructure Software Engineer Applications & iPad HW & SW. Travel req: provisioning. (Req#A28VYR) Rsrch, dsgn, dvlp 50% Hardware Development Engineer & implement innovative DSS. Mechanical Design Engineer (Req#9ZY3GD) Resp for Printed Software Engineer Applications (Req#9RXPCA) Des, dev, & de- Circuit Board (PCB) level circuit (Req#9ZY3P3) Dsgn & dvlp SW ploy automation for advncd man- dsgn & sys level intgrtion. for iTunes Store App Content QC. ufacturing processes. Travel req’d Engineering Project Lead Software Development Engineer 30%. (Req#9CSUYS) Prfrm SW & map (Req#9SGVTP) Dsgn and de- Software Engineering Project data delivery, at the highest & velop SW for embedded consumer Lead (Req#9WXS7Z) Resp for lowest lvls. devices. projct SW delivrbles, schduls & Software Engineer Applications Information Technology Project scope. (Req#9UVUCG) Des & dev sclble Specialist (Req#9U3SJA) Resp for Software Engineer Applications distrib sys. Ensure & ehnce sclblty end to end Int’l retail expansion (Req#A58NJH) Dfne & enhnce of sys. programs. cstomr inquiries supprt sys used Hardware Development Engineer Machine Learning Engineer across mltpl Apple Busnss Units. (Req#9ZZ4L3) Dsgn & dvlp inte- (Req#9T66WP) Dev innovative Software Engineer Applica- gration processes for advncd key machine lrning & data mining tions (Req#9UQT5N) Dsgn & display mod from concept to high mdls & algorithms to solve pr- dev front-end SW & usr intrfce vol prod. Travel req: 40% blms in biz apps. wirefrms for shared back-end Product Design Engineer Software Engineer (Req#9Z- infrstrctre. (Req#9T9PDB) Dev & practice K3FA) Maintain & bug fix CAD Software Development Manager analyt methds based on multiphys data mgmt w/ Perforce. (Req#9ML2SY) Mnge team of modeling & simul. Software Quality Assurance Engi- SW Dev Engrs resp for tstng Mac Engineering Project Coordina- neer (Req#9ZPV3E ) Eng systems Safari. tor (Req#9LZR7R) Dsgn & dvlp to enable QA. Software Engineer Applications SW apps for Apple’s srvce mgmnt Hardware Development Engineer (Req#A534JV) Dsgn & dev SW space to spprt Retail & Partner (Req#9F4R4F) Des & dev HW for apps. chnnl syss. Apple Macbook sys. Travel req Systems Design Engineer Software Development Engi- 25%. (Req#9QY2CM) Evaluate latest neer (Req#9L8QHR) Dsgn & Software Development Engineer iPad, iPhone & iPod HW and SW dvlp Windows Automation (Req#9GA2YH) Des & dev Blue- sys . Travel req’d 30%. Infrastructure. tooth SW & drivers for devices & Printed Circuit Board Designer Engineering Project Manager accessories. (Req#9WNVJF) Dsgn flexible & (Req#9F4T6N) Study, construct, Software Engineer Applications rigid printed circuit boards (PCB) implmnt, & sup SW solutions (Req#9X6SB3) Build Apple’s next on CAD systm for cnsmr prdcts. for Apple’s Customer Relation- gen Employee Systems platf & Software Quality Assurance En- ship Mngmnt prog in the Service suite of products gineer (Req#9W52CM) Design & space. Software Engineer Applications implmnt test infrastructure for co- Software Engineer Applications (Req#9X44KK) Design & develop dec on Apple embedded devices. (Req#9TG33S) Dsgn, build & front-end SW for shared back-end Software Engineer Applications supprt the next genration cloud infrastructure. (Req#9XXQ24) Dvlp web user pltfrm to supprt intrnt srvcs across Apple. Systems Design Engineer interface & services that au-

66 ComputingEdge August 2016 152 COMPUTER WWW.COMPUTER.ORG/COMPUTER CAREER OPPORTUNITIES CAREER OPPORTUNITIES

APPLE INC. has the following job opportunities in Cupertino, CA: Software Engineer Applications MRD) Dev verification plan for SoC HW designs. (Req#9WRUU3) Dsgn & dev da- micro-arch spec of chip dsgn Software Development Engineer tabase archtcturs for prfrmnce, compnt. (Req#9ZZ4BF) Des, dev & Im- ENGINEERING (Req#9FZNSB) Evaluate iPhone tomate app & infrastructure scalability, high availability & di- Apple Watch System Integrator plem SW for telecomm systms sastr recovry. Software Engineer Applications & iPad HW & SW. Travel req: provisioning. (Req#9CWSFS-2ND16) Suprt Systems Engineer (Req#9WNVJ2) (Req#A28VYR) Rsrch, dsgn, dvlp 50% Hardware Development Engineer Software Engineer Applications des, construction, & validation of Monitor Internal & External sys. & implement innovative DSS. Mechanical Design Engineer (Req#9ZY3GD) Resp for Printed (Req#9P34AS) Build & maintn board/system level electronics for Bld tools & automtn using SAN, highly scalable, low-latency, dist consumer electronic devices. Software Engineer Applications (Req#9RXPCA) Des, dev, & de- Circuit Board (PCB) level circuit RAID, filesystem, & IP ntwrkng Enterprise platforms (Req#9ZY3P3) Dsgn & dvlp SW ploy automation for advncd man- dsgn & sys level intgrtion. Software Engineer Systems technolgs. for iTunes Store App Content QC. ufacturing processes. Travel req’d Engineering Project Lead Software Engineer Applications (Req#9EHSXC) Research, des, Software Development Engineer (Req#9ZM2D8) Des & dev SW impl, & test storage subsys in OS Software Development Engineer 30%. (Req#9CSUYS) Prfrm SW & map (Req#9XM3BY) Des debug, & for Apple iCloud Srvces. that support various storage-rel (Req#9SGVTP) Dsgn and de- Software Engineering Project data delivery, at the highest & bring-up next gen iOS dvcs. tech, incl kernel device drivers velop SW for embedded consumer Lead (Req#9WXS7Z) Resp for lowest lvls. Software Engineer Applications Software Quality Assurance En- & user level frmwrks. Travel req devices. projct SW delivrbles, schduls & Software Engineer Applications (Req#9U3RQM) Build web- gineer (Req#9KDT6D) Resp for based, self-serv solutions to provd 15%. Information Technology Project scope. (Req#9UVUCG) Des & dev sclble testing key featres of one or more automat provisioning of load-bal- Firmware Engineer (Req#- Specialist (Req#9U3SJA) Resp for Software Engineer Applications distrib sys. Ensure & ehnce sclblty prodcts & SW updates. ancers & frewlls rel to datacenter 9FUQT3) Use anlysis or atmtd end to end Int’l retail expansion (Req#A58NJH) Dfne & enhnce of sys. Product Design Engineer mgmt. learning to dvlp & tune algrthms programs. cstomr inquiries supprt sys used Hardware Development Engineer (Req#9GC54P) Dsgn, dev, & Software Development Engineer for prcssing sgnls & optmzing Machine Learning Engineer across mltpl Apple Busnss Units. (Req#9ZZ4L3) Dsgn & dvlp inte- manufctre Printed Circuit Board (Req#A2ZVLB) Des, test, & im- prfmnce across a lrge user base. (Req#9T66WP) Dev innovative Software Engineer Applica- gration processes for advncd key (PCB) & Flexible Printed Circuit plmnt lrg-scale data pipelines & ASIC Design Engineer machine lrning & data mining tions (Req#9UQT5N) Dsgn & display mod from concept to high Assemblies (FPCA). Travel Req’d retrieval sys. (Req#9LE2DY) Rspnsbl for phys- mdls & algorithms to solve pr- dev front-end SW & usr intrfce vol prod. Travel req: 40% 30% ical dsgn & implementation of blms in biz apps. wirefrms for shared back-end Product Design Engineer Software Development En- gineer (Multiple Positions) partitions. Software Engineer (Req#9Z- infrstrctre. (Req#9T9PDB) Dev & practice Apple Inc. has (Req#A5G2MX) Dsgn & imple- Hardware Development Engineer K3FA) Maintain & bug fix CAD Software Development Manager analyt methds based on multiphys ment production algorithms that (Req#9T7Q3D) Des & dev mech the following job data mgmt w/ Perforce. (Req#9ML2SY) Mnge team of modeling & simul. process text & other data. test fixtures to support prod dev opportunities in Software Quality Assurance Engi- SW Dev Engrs resp for tstng Mac Engineering Project Coordina- Software Engineer Systems & manufact. Travel req 25%. neer (Req#9ZPV3E ) Eng systems Safari. tor (Req#9LZR7R) Dsgn & dvlp Orlando, FL: (Req#9XQQAE) Supp prod dev. Software Engineer Systems to enable QA. Software Engineer Applications SW apps for Apple’s srvce mgmnt ASIC Design Engineer (Req#- (Req#A534JV) Dsgn & dev SW space to spprt Retail & Partner Ident probs & verify sols thru alg (Req#9T64M9) Integrate, deploy, Hardware Development Engineer des for Apl prods. & test intelligent algorithmic sys. 9FLV3V) Dev SW to validate (Req#9F4R4F) Des & dev HW for apps. chnnl syss. ASIC Design Engineer Hardware Development Engineer functionality of Apple’s mobile Apple Macbook sys. Travel req Systems Design Engineer Software Development Engi- (Req#9HMQHG) Resp for ver- (Req#9YTNBM) Dsgn, dev & GPU HW. 25%. (Req#9QY2CM) Evaluate latest neer (Req#9L8QHR) Dsgn & ifcation of multiple intellectual validate radio baseband HW cir- Software Development Engineer iPad, iPhone & iPod HW and SW dvlp Windows Automation sys . Travel req’d 30%. Infrastructure. prop blocks on Apple’s System- cuits & sys for wireless commu- (Req#9GA2YH) Des & dev Blue- on-a-Chip (SoC) HW dsgns. nications devics. Travel req’d 15% tooth SW & drivers for devices & Printed Circuit Board Designer Engineering Project Manager Systems Design Engineer (Req#- Hardware Development Engi- accessories. (Req#9WNVJF) Dsgn flexible & (Req#9F4T6N) Study, construct, rigid printed circuit boards (PCB) implmnt, & sup SW solutions 9FLQCW) Eval Active OTA Per- neer (Req#9T8VMC) Dsgn & dev Software Engineer Applications formance. Characterize passive/ touch HW solutions. (Req#9X6SB3) Build Apple’s next on CAD systm for cnsmr prdcts. for Apple’s Customer Relation- active antenna perform incl effi- Software Engineer Systems (Req#- gen Employee Systems platf & Software Quality Assurance En- ship Mngmnt prog in the Service space. ciency, gain, & pattrn. 9STUDQ) Dsgn & produce suite of products gineer (Req#9W52CM) Design & Refer to Req# & mail resume implmnt test infrastructure for co- Software Engineer Applications Software Development Engineer highly integrated SW systms for Software Engineer Applications (Req#9JFPY8) Res for the des new prdcts. to Apple Inc., ATTN: L.J., 1 (Req#9X44KK) Design & develop dec on Apple embedded devices. (Req#9TG33S) Dsgn, build & Infinite Loop 104-1GM, Cu- & dev of SW for the Siri server ASIC Design Engineer front-end SW for shared back-end Software Engineer Applications supprt the next genration cloud pertino, CA 95014. Apple is an platform. (Req#9QLUDA) Verify intellec- infrastructure. (Req#9XXQ24) Dvlp web user pltfrm to supprt intrnt srvcs EOE/AA m/f/disability/vets. across Apple. ASIC Design Engineer (Req#9NS- tual property blocks on Apple’s Systems Design Engineer interface & services that au-

www.computer.org/computingedge 67 152 COMPUTER WWW.COMPUTER.ORG/COMPUTER AUGUST 2016 153 TECHNICAL Oracle America, Inc. CONSULTING Oracle America, Inc. has openings for Oracle America, Inc. has openings for SENIOR LICENSE has openings for CONSULTANT CONSULTANT TECHNICAL positions in positions in ANALYST Burlington, MA. Minneapolis, MN. Job duties include: Perform software Job duties include: Analyze require- positions in Orlando, FL. license reviews of our partners to ensure ments and deliver functional and compliance with their contract terms and Job duties include: Analyze user technical solutions. Implement Oracle’s Business Practices while providing products and technologies to meet requirements to develop, implement, value-added services to enhance partner post-sale customer needs. Travel to and/or support Oracle’s global satisfaction. Travel to various, unantici- infrastructure. pated sites throughout the United States various unanticipated sites throughout required. the United States required. Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], [email protected], referencing 385.19577. referencing 385.17121. referencing 385.19352. Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity.

SOFTWARE Oracle America, Inc. Ooyala Inc. has openings for has openings in our Santa Clara, CA location for: Software Engineer (6688.96) Design & deliver solutions to improve workflows SOFTWARE across multiple teams; Software Engineer (6688.119) Design & deliver innovative solutions in the form of high-quality, bandwidth-efficient APIs; Software Develop- DEVELOPER ment Engineer In Test - Big Data (6688.113) Implement framework & automate positions in Naples, FL. testing scripts in collaboration with software development for software enhancement & new products; Software Engineering in Test (6688.116) Review new Job duties include: Design, feature requirements & designs, & work closely with team to define strategy for develop, troubleshoot and/or testing various components. test/QA software.

Ooyala Inc. has openings in our Plano, TX location for: Software Development Engineer in Test (6688.86) Develop tests & build & Apply by e-mailing resume to maintain automated test infrastructure. [email protected] , referencing 385.18278. Send resume to HR, 4750 Patrick Henry Drive, Santa Clara, CA 95054. Must ref. job code above when applying. Oracle supports workforce diversity.

68 ComputingEdge August 2016 154 COMPUTER WWW.COMPUTER.ORG/COMPUTER TECHNICAL CONSULTING It’s work that matters. It's what we do at Symantec. Symantec is the world leader in providing solutions to help individuals and Oracle America, Inc. enterprises assure the security, availability, and integrity of their information. In essence, we protect the free flow of information in a Oracle America, Inc. has openings for Oracle America, Inc. connected world. As the fourth largest independent software company in the world, Symantec has operations in more than 40 countries has openings for has openings for SENIOR LICENSE with 475 out of Fortune's global 500 companies using our solutions. People look to us to safeguard the integrity of their information, ensuring it is secure and available. Achieving this ambitious goal is only possible through the combined efforts of the innovators and CONSULTANT CONSULTANT visionaries that Symantec continuously attracts. Symantec draws the very best people with a variety of backgrounds, experiences and TECHNICAL positions in positions in perspectives and provides them with a work environment where uniqueness is valued and empowered. The creative people we attract Burlington, MA. Minneapolis, MN. ANALYST help define the spirit of innovation at Symantec. Symantec is proud to be an equal opportunity employer. Job duties include: Perform software Job duties include: Analyze require- positions in Orlando, FL. license reviews of our partners to ensure ments and deliver functional and Symantec Corporation currently has openings for the following positions in Cambridge, MA (various levels/types): compliance with their contract terms and technical solutions. Implement Job duties include: Analyze user Oracle’s Business Practices while providing products and technologies to meet Software QA Engineers (SQAMA716) Responsible for developing, applying and maintaining quality standards for company products. requirements to develop, implement, value-added services to enhance partner post-sale customer needs. Travel to and/or support Oracle’s global satisfaction. Travel to various, unantici- Develop and execute software test plans. Analyze and write test standards and procedures. infrastructure. pated sites throughout the United States various unanticipated sites throughout required. the United States required. Symantec Corporation currently has openings for the following positions in Culver City, CA (various levels/types): Apply by e-mailing resume to Apply by e-mailing resume to Apply by e-mailing resume to [email protected], [email protected], [email protected], Engineering Managers (EMCC716) Direct and supervise team of engineering (QA and/or development teams). Develop standards for referencing 385.19577. referencing 385.17121. referencing 385.19352. products and/or oversee development and execution of software and/or analysis of test results.

Oracle supports workforce diversity. Oracle supports workforce diversity. Oracle supports workforce diversity. Software Engineers (SWECC716) Responsible for analyzing, designing, debugging and/or modifying software; or evaluating, developing, modifying, and coding software programs to support programming needs. SOFTWARE Oracle America, Inc. UI Designers (UIDCC716) Responsible for providing User Interface (UI) design and support to product development teams, including the design, analysis and investigation of applications, systems, and Graphic User Interfaces (GUIs). Ooyala Inc. has openings for has openings in our Santa Clara, CA location for: Symantec Corporation currently has openings for the following positions in Herndon, VA (various levels/types): Software Engineer (6688.96) Design & deliver solutions to improve workflows SOFTWARE across multiple teams; Software Engineer (6688.119) Design & deliver innovative Security Infrastructure Administrators (SIAVA716) Assist with security infrastructure administration as part of Symantec’s Network solutions in the form of high-quality, bandwidth-efficient APIs; Software Develop- DEVELOPER Operations team. ment Engineer In Test - Big Data (6688.113) Implement framework & automate positions in Naples, FL. testing scripts in collaboration with software development for software enhance- Software Engineers (SWEVA716) Responsible for analyzing, designing, debugging and/or modifying software; or evaluating, development & new products; Software Engineering in Test (6688.116) Review new Job duties include: Design, ing, modifying, and coding software programs to support programming needs. feature requirements & designs, & work closely with team to define strategy for develop, troubleshoot and/or Submit resume to [email protected] . Must reference position & code listed above. EOE. For additional testing various components. test/QA software. information about Symantec and other positions visit our website at http://www.symantec.com. Ooyala Inc. has openings in our Plano, TX location for:

Software Development Engineer in Test (6688.86) Develop tests & build & Apply by e-mailing resume to maintain automated test infrastructure. [email protected] , referencing 385.18278. Send resume to HR, 4750 Patrick Henry Drive, Santa Clara, CA 95054. Must ref. job code above when applying. Oracle supports workforce diversity.

www.computer.org/computingedge AUGUST 2016 15569 154 COMPUTER WWW.COMPUTER.ORG/COMPUTER CLOUD TIDBITS

of networking into one that’s more software-defined and on-demand. Software- In this article, I’ll review the advantages of SDN, describe how to exploit it, and explain related business and technology opportunities that might not readily be clear. Moreover, I’ll look at the potential for Defined SDN solutions to live in the cloud, providing an on- demand networking capability that could offer better and cheaper approaches to networking, and the abil- Networks ity to place network volatility into its own domain. SDN Meets Cloud SDN is the software enablement of infrastructures. It’s part of the software-defined datacenter concept Meet Cloud that, in essence, detaches the hardware from the software, and replaces complex hardware devices with changeable layers of software. Specifically, SDN decouples the software control panel from a Computing node’s forwarding hardware (such as routers and switches), and executes the control software in either a local server or the cloud. A couple of things are behind the shift to SDN. SOFTWARE-DEFINED NETWORKING (SDN) First, as enterprises relocate their workloads, in- REMOVES THE LIMITATIONS OF HARDWARE cluding data and applications, to public clouds, the FOR THOSE WHO BUILD NETWORK INFRA- center of the workload universe is shifting, and the STRUCTURES. For example, SDN adds the ability network must follow. Second, as net-new applica- to change the networking hardware’s behavior on tions appear on public clouds, networks need to be the fly. Although this seems like a good idea for on- on clouds as well. premises systems, there’s a huge advantage associ- When considering whether to deploy an SDN ated with employing SDN in the cloud as well. solution in the cloud, your decision will come down However, the use of SDN within public and to its orchestration capabilities. Indeed, effective private clouds as well as on the edge of clouds goes orchestration is the key factor for success when beyond its typical application. Indeed, this shift has dealing with a complex infrastructure that executes the potential to turn the hardware-driven concept millions of software transactions, and thus can link an SDN solution directly with applications. This ability will likely be a key benefit and driver of SDN, and SDN in the cloud. There are several key value points to using SDN and the cloud. First, it lets you software-define everything, so you can run the software on traditional systems as well as private and public clouds. Second, it lets you couple the network requirements of the DAVID S. applications or workloads to the network configu- LINTHICUM ration. You can configure the network for whatever purpose, on the fly. The objective is to eliminate the Cloud Technology Partners need to find a common configuration that works [email protected] well with most applications, but each configuration carries with it the ability to define network require-

d3tid.indd 8 7/12/16 1:05 PM CLOUD TIDBITS

of networking into one that’s more software-defined ments on the fly. Essentially, in an SDN to install patch after patch. Many of ture of SDN, the network itself can spot and on-demand. solution, control is directly program- the recent breaches have cost compa- repeating attack patterns, not just a Software- In this article, I’ll review the advantages of SDN, mable. Thus, administrators have more nies millions of dollars in lawsuit pay- single instance. Thus, the SDN solution describe how to exploit it, and explain related busi- power than ever before, which is a huge outs, as well as bad public relations, and can provide dynamic defenses without ness and technology opportunities that might not step forward from today’s static routing. they were all traced back to bad prac- having to shut down the target applica- readily be clear. Moreover, I’ll look at the potential for Static routing isn’t application depen- tices in updating network switches and tions and databases. Users should be Defined SDN solutions to live in the cloud, providing an on- dent, and can be nearly impossible to hubs to protect against new but known unaware of the attack. demand networking capability that could offer better change on the fly to dynamically adapt vulnerabilities. Governance is also much easier to and cheaper approaches to networking, and the abil- to an application’s specific needs. This Finally, SDN offers elastic capac- deal with, since we can dynamically ity to place network volatility into its own domain. has been called network as code. ity. Although this is true when an SDN place policies on network functions, and Networks Another key value point is the solution runs on owned hardware, the even integrate the network governance SDN Meets Cloud ability to manage the network as soft- larger benefit happens when SDN so- with other governance layers, such as SDN is the software enablement of infrastructures. ware, enabling self-healing capabilities lutions run in public clouds. Public services/APIs and data. This means that It’s part of the software-defined datacenter concept via software updates. You can exploit clouds provide auto- and self-scaling governance systems can work together Meet Cloud that, in essence, detaches the hardware from the SDN’s dynamic nature and quickly capabilities, meaning that your SDN so- with tight integration between securi- software, and replaces complex hardware devices change network behaviors. Using SDN lution can allocate more logical servers ty, data, applications, services, and the with changeable layers of software. Specifically, and the cloud also allows you to link the to support an increasing network load. network to ensure that those who use SDN decouples the software control panel from a network with DevOps processes, letting You won’t have to worry about your net- the network and the network services Computing node’s forwarding hardware (such as routers and you, for example, test applications un- work’s scalability as long as the SDN are authorized, and that all proper limi- switches), and executes the control software in ei- der different networking configurations services run on a public cloud platform. tations are imposed. ther a local server or the cloud. that are dynamically defined. However, in a public cloud, this is only A couple of things are behind the shift to SDN. The capital versus operating expen- true at the compute/infrastructure lay- What’s Next? SOFTWARE-DEFINED NETWORKING (SDN) First, as enterprises relocate their workloads, in- diture thing is alive and well in SDN. er, not at the network layer. SDN is a new option for updating and REMOVES THE LIMITATIONS OF HARDWARE cluding data and applications, to public clouds, the SDN allows many of the services that modernizing your on-premises network. FOR THOSE WHO BUILD NETWORK INFRA- center of the workload universe is shifting, and the were once delivered by hardware to be- Security and Governance It also lets you leverage the public cloud STRUCTURES. For example, SDN adds the ability network must follow. Second, as net-new applica- come software-delivered services. Enter- Advantages to make your SDN solution that much to change the networking hardware’s behavior on tions appear on public clouds, networks need to be prises are sick and tired of having to buy Public clouds are centrally managed by more responsive and inexpensive to op- the fly. Although this seems like a good idea for on- on clouds as well. new network devices every few years. an operations staff that can be aware of erate. This both revolutionizes the way premises systems, there’s a huge advantage associ- When considering whether to deploy an SDN SDN removes the networking function security issues and update the SDN solu- we do networking and makes the net- ated with employing SDN in the cloud as well. solution in the cloud, your decision will come down from dependence on any device; mov- tions running on the cloud accordingly. working solutions application-ware. However, the use of SDN within public and to its orchestration capabilities. Indeed, effective ing SDN into the cloud can save an en- This removes the enterprise network se- The dynamic linking of applications private clouds as well as on the edge of clouds goes orchestration is the key factor for success when terprise even more money and increase curity administrator from having to deal using SDN has so many advantages that beyond its typical application. Indeed, this shift has dealing with a complex infrastructure that executes efficiency. However, keep in mind that with an endless number of threats and it’s impossible to ignore the value. Un- the potential to turn the hardware-driven concept millions of software transactions, and thus can link the public cloud already implements its vulnerabilities that require constantly like other new software-defined innova- an SDN solution directly with applications. This own form of SDN by abstracting a few stopping and updating the network. tions that come with varying degrees of ability will likely be a key benefit and driver of SDN, networking layers and providing simple In other words, the network itself can return on investment, SDN in the cloud and SDN in the cloud. interfaces to manipulate it. The actual change behavior to stop attacks, rather will be a win for any enterprise that There are several key value points to using SDN routing and switching of packets in a than wait for external systems to come adopts it. However, like anything else, and the cloud. First, it lets you software-define ev- public cloud is never exposed as a con- into play to block IP addresses or shut it requires some deep understanding, erything, so you can run the software on traditional figurable option for customers. Thus, down resources that are being attacked. thinking, and planning. systems as well as private and public clouds. Second, although some features are relocatable, For example, say our application Core to the new SDN vision is that it lets you couple the network requirements of the others aren’t. and data are under a distributed denial SDN solutions will live at different lay- DAVID S. applications or workloads to the network configu- SDN also provides better and more of service attack. When the security ers within a cloud-enabled enterprise. LINTHICUM ration. You can configure the network for whatever dynamic security and governance. As systems spot the attack, the network In layer 1, applications will be able purpose, on the fly. The objective is to eliminate the security threats are understood, the could lock out the offending IP address to define SDN behaviors with code. Al- Cloud Technology Partners need to find a common configuration that works SDN solution can dynamically update or shut down resources. However, IP though the SDN solution might not live [email protected] well with most applications, but each configuration itself to protect against those threats. addresses can change using the same within the applications, it will be driven by carries with it the ability to define network require- Security administrators no longer need attack patterns. Given the dynamic na- the applications. Dynamic configuration

d3tid.indd 8 7/12/16 1:05 PM d3tid.indd 9 7/12/16 1:05 PM CLOUD TIDBITS This article originally appeared in IEEE Cloud Computing, vol. 3, no. 3, 2016.

changes can occur as needed, and the net- proach, where all networks, as well as a solution where most of the processing work will reconfigure around the applica- other infrastructure, can be managed occurs in the cloud. In some instances, tion, not the other way around. and monitored using a single dashboard cloud providers will offer networking At the second layer, data has a linked and set of controls. as a service. You’ll be able to take ad- relationship with the SDN solution as Layer 6 concerns security, which is vantage of these services if your prob- well. Data is coupled to the application, also systemic to an SDN configuration. lem domain is simple and cloud-centric. so all the rules in layer 1 apply to layer Layering security directly into the SDN However, most of us have complex work- 2. Indeed, the data can dynamically de- solution provides dynamic and extend- loads that will be distributed among fine the network’s behavior. For example, able defenses for the infrastructure and traditional systems, private clouds, and an SDN can dynamically increase band- workloads, including the ability to spot public clouds. For this complex prob- width for the application and data in- patterns based on dynamic monitoring. lem domain, SDN solutions must be stance to accommodate a huge result set It might also include learning system defined, implemented, and maintained coming from the database. capabilities, such as the use of machine by network administrators that have At layer 3, the existing on-premises learning. shifted to the use of a distributed cloud infrastructure needs to provide the SDN and a noncloud model, which will need solution with a set of resources to per- Mapping Your Own Path to be custom integrated. form operations on-premises. This Challenges to SDN and the cloud ex- These problems are easily over- includes services and traditional network- ist. First, it won’t protect you from bad come. The easy bet to place is that pub- ing equipment that still serve a purpose. software updates. Indeed, whereas tra- lic cloud providers will offer their own However, this equipment isn’t subservi- ditional networking can resume with SDN services, as well as learn how to ent to the SDN solution, which might just a power reset, SDN requires major work and play with the popular solu- live in the cloud, on-premises, or both. surgery to recover from a break in ser- tions that enterprises leverage. Activities at layer 4 will depend on vice. Second, you need to manage the whether the SDN solution lives in a pri- configurations and policies, so you need vate or public cloud environment. As- to add SDN configuration to the list of THE LARGE VALUE HERE IS THAT suming that most enterprises deploying things you need to track. WE NO LONGER HAVE TO ACCEPT SDN solutions will select public clouds, Best practices are beginning to NETWORKS AS STATIC THINGS. the core SDN processing and storage will emerge around SDN, as well as around With the new software-defined capabili- take place on that public cloud. However, the use of SDN with clouds and on ties, networks can change around work- network latency over the open Internet clouds. They include separating out the load needs, which can change around could prove problematic for a remotely network planning from the application the business’s needs. This makes our hosted SDN solution. Thus, SDN will design and requirements processes. and networks much more effective and ef- typically live both on a public cloud and placing layers of governance and security ficient, and no longer coupled to hard- on-premises, with the public cloud pro- within the SDN solution, and software- ware. That’s the name of the game viding the master control as well as soft- defining them along with the network. here. ware updates. The network’s processing The largest obstacle in SDN isn’t loads can then be distributed between the technology, but the leap of faith that private and public cloud instances to pro- network administrators must take as DAVID S. LINTHICUM is senior vice mote stability and higher performance. networks shift from being hardware- president of Cloud Technology Partners. Layer 5 involves management and defined to being software-defined. Right He’s also Gigaom’s research analyst and monitoring, which are systemic to SDN. now, SDN solutions are largely used for frequently writes for InfoWorld on deep Management and monitoring tools net-new systems and infrastructure, and technology subjects. His research inter- will ensure that the SDN solution pro- they’re often leveraged improperly. For ests include complex distributed systems, vides adequate support for the applica- instance, many don’t leverage the dy- including cloud computing, data inte- tion workloads and user engagement, namic capabilities of SDN or use SDN gration, service-oriented architecture, as well as for established service-level with emerging cloud-based systems. Internet of Things, and big data systems. agreements. These tools should be ag- As workloads shift from on-premises Contact him at david@davidlinthicum gregated into a “single pane of glass” ap- to the cloud, the clear path to SDN is .com.

72 ComputingEdge August 2016 10 IEEE CLOUD COMPUTING WWW.COMPUTER.ORG/CLOUDCOMPUTING

d3tid.indd 10 7/12/16 1:05 PM 2017 Richard E. Merwin Award for Distinguished Service CALL FOR AWARD NOMINATIONS Deadline 15 October 2016

ABOUT THE MERWIN AWARD The highest level volunteer service award of the IEEE Computer Society for outstanding service to the profession at large, including significant service to the IEEE Computer Society or its predecessor organizations. ABOUT RICHARD MERWIN Richard Merwin was a pioneer in digital computer engineering who participated in the development of the ENIAC, MANIAC, and STRETCH computers. Despite a busy and productive technical career, Merwin found time to be active in professional MICHAEL R. WILLIAMS societies, including the IEEE Computer Society, 2016 Richard E. Merwin Award ACM and AFIPS. His generosity of spirit and genuine for Distinguished Service helpfulness was an important element in the progress of the computer profession. In recognition of exemplary contributions to the computing AWARD profession through outstanding A bronze medal and $5,000 honorarium are awarded. scholarship and enduring and committed volunteer PRESENTATION participation and leadership of The Richard E. Merwin Award is presented at the the IEEE Computer Society. IEEE Computer Society’s Annual Awards Ceremony. REQUIREMENTS This award requires 3 endorsements. AWARDS HOMEPAGE NOMINATION SUBMISSION www.computer.org/awards Nominations are being accepted electronically www.computer.org/web/awards/merwin CONTACT US [email protected] OF CYBERSECURITY AND THREATS COUNTER MEASURES

Learn How You Can Develop Real-World Security Solutions That Work For Your Organization

13 September 2016 | Seattle, Washington

What Do You Need to Know About Rock Star Speakers Cybersecurity?

Lots!

The attackers have gotten more sophisticated. No company or person is safe. At Rock Stars of Cybersecurity, we’ve brought together the real leaders in this critical technology – Google, Adobe, PayPal, Intel Health and Life Sciences, and others – to talk about the trends, cybersecurity programs and advice on how you can develop real-world Drew Hintz Bruce Snell Hui Wang security solutions that work for your organization – Vulnerability Analyst Cybersecurity and Sr. Director of Global and that don’t disrupt your operations. Google Privacy Director, Risk Sciences Intel Security Paypal