Professor Juho Kim＇s research team received ACM CS..

Professor Juho Kim's research team received ACM CSCW 2019 Best Paper Honorable Mention Award Professor Joho Kim and his research team were awarded Best Paper Honorable Mention Award at ACM CSCW 2019 (The 22nd ACM Conference on Computer-Supported Cooperative Work and Social Computing) in Austin, America on December 9~13, 2019. The ACM CSCW Society is one of the highest authority in the field of human-computer interaction, and the Best Paper Honorable Mention Award is awarded to the top 5％ of all papers. As a result of international cooperation, this study was expanded by John Joon Young Chung (first author, currently a Ph.D. student at the University of Michigan), who had worked as a researcher at Professor Kim's lab. Paper title: ‘Efficient Elicitation Approaches to Estimate Collective Crowd Answers’ In this paper, as a study of crowdsourcing techniques that are essential for generating datasets for machine learning, the authors introduced techniques that can be applied to subjective and ambiguous tasks, such as emotional recognition without clear answers, or identifying relationships between words. Experimental comparisons of the different questionnaires provided to crowd workers showed that the most efficient case was. This method required 40％ less manpower while reducing 21.4％ of the work time compared to the basic method. This work is meaningful in that it introduced a technique that enables substantial cost savings in building a variety of machine learning-based AI applications that require large datasets. Blog post: https://medium.com/acm-cscw/efficient-elicitation-approaches-to-estimate-collective-crowd-answers-bd4c9adddb18 Paper link: https://dl.acm.org/citation.cfm?id=3359164 Paper information: John Joon Young Chung, Jean Y. Song, Sindhu Kutty, Sungsoo (Ray) Hong, Juho Kim, and Walter S. Lasecki. 2019. Efficient Elicitation Approaches to Estimate Collective Crowd Answers. Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 62 (November 2019), 25 pages. DOI: https://doi.org/10.1145/3359164

KAIST and MarkAny developed an AI-based content in..

A research team led by Prof. Heung­Kyu Lee in the KAIST computer science department together with MarkAny corp. developed an AI-based content integrity inspection service, which verifies forgery and falsification of digital images and videos. Articles: ETNews: http://www.etnews.com/20191029000081 The Chosun Ilbo: http://it.chosun.com/site/data/html_dir/2019/10/29/2019102900980.html

Prof. Min H. Kim delivered a keynote lecture at CA..

Prof. Min H. Kim delivered a keynote lecture at CAD/Graphics 2019 ​ Prof. Min H. Kim delivered a keynote lecture on the topic “Beyond Human Vision - Seeing More with Camera” at Computer-Aided Design and Computer Graphics (CAD/Graphics) 2019 in Qingdao, China, from May 5 to 6. Prof. Kim discussed the foundations of compressive spectral imaging and introduced recent advances in the development of hyperspectral imaging technology that has been developed to make spectral imaging more accurate and smaller for practical applications in our daily life.

Prof. Sung Yong Shin received Lifetime Achievement..

Prof. Sung Yong Shin, awarded Asia Graphics Lifetime Achievement Award. Prof. Sung Yong Shin received an Asia Graphics Lifetime Achievement Award in Pacific Graphics 2019 for his outstanding leadership and tremendous contributions to computer graphics research in Asia. Prof. Shin founded Pacific Graphics in 1993 with his companion Prof. Myoung Soo Kim at Seoul National University and Prof. Shin is an emeritus professor of computer science at the KAIST School of Computing.

KAIST, codeveloped an AI-education program with Lu..

Prof. Sungho Jo established an AI-education program with Luxrobo (CEO: Sanghun Oh) - an edu-tech startup in South Korea. DETAILS

Editorial board member of ACM Transactions on Soft..

Dr. Shin Yoo, a faculty member of the School of Computing, KAIST, has been appointed as an editorial board member of ACM Transactions on Software Engineering and Methodology (TOSEM) starting from July 2019. TOSEM is one of two SCI journals in software engineering, along with IEEE Transactions on Software Engineering. TOSEM has an impact factor of 2.516 in JCR 2017 and is currently ranked at the 18th place by Google Scholar among the Software Systems conferences and journals.

Prof. Yoon organized a workshop on sound source lo..

Prof. Yoon organized a workshop on sound source localization for robots at ICRA 2019, which is the biggest robotics conference. https://sglab.kaist.ac.kr/SSL_ICRA19/index.html While its research community is not big yet, Prof. Yoon brought related researchers together to make a bigger one.

ICSE 2019 ACM SIGSOFT Distinguished Paper Award

Together with Dr. Kihong Heo (Univ. of Pennsylvania) and Prof Hakjoo Oh (Korea Univ.), Prof Hongseok Yang at KAIST received the ICSE 2019 ACM SIGSOFT Distinguished Paper Award for their paper “Resource-aware Program Analysis via Online Abstraction Coarsening.” ICSE is a top international conference in software engineering.

Posted research from School of Computing at KAIST ..

Research by Professor Cha Mi-young and Professor Shin In-shik of School of Computing were posted in the "KAIST Breakthroughs" (Spring 2019), which is an english study webzine of the College of Engineering. Professor Cha Mi-young ＊ Title : Using deep learning to detect fake news online ＊ Link : http://breakthroughs.kaist.ac.kr/?post_no=1318 Professor Shin In-shik ＊ Title : UbiTap: Leveraging acoustic dispersion for ubiquitous touch interface on solid surfaces ＊ Link : http://breakthroughs.kaist.ac.kr/?post_no=1317

Graduate School of Information Security published ..

KAIST's Graduate School of Information Security will publish three papers submitted to the International Society of Network and Distributed System Security (NDSS), one of the world's four major conferences of security, held in San Diego, U.S. from February 24 to 27. NDSS is one of the world's top four academic societies in international security, along with ACM CCS(Conference on Computer and Communications Security), USENIX Security, and IEEE S&P (Security and Privacy). In particular, only 89 of the 521 papers submitted this year were published, and the adoption rate of theses was only 17 percent, adding to the significance. [Cybercriminal Minds: An invasive study of cryptocurrency abusers in the Dark Web] Lee Seung-hyun, Yoon Chang-hoon, Kang Hee-do, Kim Yeon-geun (Ph.D students), and Kim Yong-dae, Han Dong-soo, Son Sooel, and Shin Seung-won (Professors), a research team from Graduate School of Information Security, investigated the scale and characteristics of illegal transactions based on virtual money on the dark web. [CodeAlchemaist: Semantics-Aware Code Generation to Find Vulnerabilities in JavaScript Engines] Han Hyung-seok in Ph.D. course, Oh Dong-hyun in Master course, and Professor Cha Sang-gil at the Graduate School of Information Security, described the technology in this paper to automatically find vulnerabilities present in JavaScript engines used in web browsers. [OBFSCURO: A Commodity Objectification Engine on Intel SGX] Cho Byung-gil in School of Computing Ph.D. course, Shin In-sik, a professor of Graduate School of Information Security, and Lee Byung-young, a professor at Seoul National University, proposed in this paper a program obfuscation technology to enhance security on the SGX environment.

Professor Kwang-jo Kim selected as Editor of IEEE ..

Kim Kwang-jo, a professor at the Graduate School of Information Security of School of Computing, was chosen as an editor of IEEE Trans. on Dependable and Secure Computing, an international journal on information protection. This journal is one of 17 journals organized by the IEEE Computing Society, which has an impact factor of 2.926 and H-Index of 56, which bimonthly publishes the latest research results on basic theories, design methodologies, and mechanisms on design, modeling, and evaluation techniques for systems and networks that provide reliable security. Professor Kim has been selected as the first generation cryptographer in Korea who has been engaged in research on information protection and cipher since 1982, and is currently a member of the IACR (International Society for Cryptography) stone academy for his excellent research performance and international reputation. [Related Web Page] https://www.computer.org/web/tdsc/about Aims & Scope of the Journal IEEE Transactions on Dependable and Secure Computing (TDSC) publishes archival research results focusing on research into foundations, methodologies, and mechanisms that support the achievement–through design, modeling, and evaluation–of systems and networks that are dependable and secure to the desired degree without compromising performance. The focus also includes measurement, modeling, and simulation techniques, and foundations for jointly evaluating, verifying, and designing for performance, security, and dependability constraints. The IEEE Computer Society sponsors more than 200 technical conferences and events each year, including the industry-oriented "Rock Stars" series, all over the world, aimed at research and industry professionals. Our publications are peer-reviewed, indexed, and authored by technology thought leaders worldwide, and include 17 scholarly journals and 13 magazines featuring the latest technology trends

˝Move it my way！˝ ： Development of Robot Hands to ..

This is a press article from Professor Cho Seong-ho of KAIST School of Computing in connection with the development of wearable robots that help people who cannot use their hands due to disabilities. https://ytn.co.kr/_ln/0105_201902010236590670 http://news.kbs.co.kr/news/view.do?ncd=4130095&ref=A

Professor Moonzoo Kim, adopted an ICSE 2019 SE in ..

ACM/IEEE ICSE Software Engineering in Practice (SEiP) track, the world's top academic society in software engineering, has adopted a paper by Professor Kim Moon-joo of our Computing Department. This paper, "Automaticized Test Generation to Achieve High Test Coverage with Reduced Human Effort in the Automotive Industry," is a research paper written by Kim Yoon-ho and Moonzoo Kim of the Department of Computer Science. The developed and applied MIST, a SW auto-test tool for vehicles, automatically tested SW (Integrated Body Unit) for modern Mobis vehicles to achieve 90％ of the quarterly coverage and reduced the time and cost of SW testing to less than half. The research results were published in major media outlets such as Chosun Ilbo, Maeil Economic Co., and Yonhap News Agency. https://m.yna.co.kr/view/AKR20180720158800003 http://biz.chosun.com/site/data/html_dir/2018/07/22/2018072200441.html http://news.mk.co.kr/newsRead.php?year=2018&no=460304

Presented a paper at ACM POPL 2019, the world＇s le..

ACM POPL 2019, one of the world's top academic societies in programming languages, was adopted in a paper written by KAIST Computer Science Master's Degree Park Kyung-hee, Bachelor's Degree Hong Jae-min and Professor Ryu Seok-young with Dr. Guy L. Steele Jr. of the Oracle Institute.

MARS AI Integrated Research Center hosts ‘KAIST AT..

In NeurlPS, one of the most notable international societies in machine learning, MARS AI Integrated Research Center and our school held ‘KAIST AT NeuralPS 2018’. The event was organized to promote our academic achievements and exchange with overseas researchers of artificial intelligence. MARS AI Integrated Research Center, which is a part of our school and the School of Electrical and Electronic Engineering, held an event called ‘KAIST AT NeuralPS 2018’ to promote the research and interact with the leading international artificial intelligence researchers. In addition to our school, more than 100 Korean machine learning researchers attended overseas universities such as MIT, Oxford, Univ. of Michigan, and Univ. of Toronto and domestic universities such as Seoul National University and Yonsei University and worked in industries such as Google, Naver, NVIDIA. The NeurlPS 2018 conference, which was held in Montreal, Canada, was attended by researchers from around the world and presented more than 1,000 papers. Our school presented six papers, one of which earned the honor of Spotlight. A detailed introduction to the papers and events published in NeurlPS 2018 can be found on the following website: http://ml.kaist.ac.kr/neurips2018

KAIST Develops Next Generation Binary Analysis Pla..

Over the past three years, Cha Sang-gil, a professor of School of Computing of KAIST, and his students have developed B2-R2, a technique for analyzing vulnerabilities in binary code using artificial intelligence technology with Cybersecurity Research Center(Director Professor Kim Yong-dae of Electrical and Electronic Engineering). B2-R2 is a performance system of 'Automatic Reverse Engineering and Vulnerability Detection-Based Technology Development' that is a research project by the Ministry of Science and ICT that can be used in various computer security areas such as ▲software security vulnerability analysis ▲Analysis of malicious code ▲Releasing obfuscation ▲Security Patch ▲Automatic Creation of Exposure plot. Compared to the binary analysis systems developed overseas, B2-R2 system is highly utilized (available on all operating systems such as Windows, Linux, Mac, Android, iOS) and has at least 2 to 100 times the analysis speed and maximizes the ease of analysis using the latest functional language, F#, and has a variety of interfaces with 32 programming languages. Not only is it the first binary code analysis system in Korea, but it also surpasses the BAP of Carnegie Mellon University and Angr of UCSB, which lead the field. The importance of binary analysis technology, known through ‘Next Generation binary analysis platform B2-R2 Technology Explanation Conference’ on Dec. 4, has become widely known through the Cyber Grand Challenge (CGC), the world's first computer-to-computer hacking attack and defense competition hosted by the Defense Department of Defense (DARPA) in 2016. With various security threats such as malicious code, APT attack, and DDoS attack being carried out by human hackers, the competition has become a signal for the emergence of AI-based machine hackers and can be understood as a core technology to prepare for a paradigm shift for information protection of the fourth industrial revolution. Professor Cha Sang-gil is known to be the main author who designed the core engine of the CGC championship, Mayhem, which was held in the U.S. and won the title by overwhelming margin at 2018 Information Security R&D Data Challenge Auto Detect based on AI, a Korean version of the CGC, on November 30. It is known that at the center of the system used for the contest by Cha Sang-gil B2-R2 technology which he developed with KAIST students for three years. According to information protection experts, KAIST's Cha Sang-gil team has secured key technologies in the field of 'automatic detection and response of AI-based vulnerabilities' that surpasses the world's highest level through this technical briefing and domestic competitions.

Undergraduate Jae-min Hong of School of Computing,..

At APLAS (Asian Symposium on Programming Language and Systems) 2018 which was held in Wellington, New Zealand on 12/3-5 and was sponsored by KAIST, Hong Jae-min (Advisor Ryu Seok-young) received the second prize for his thesis on Path Dependent Ties with Path-Equality at the Student Research Competition, which he participated with graduate students. He published the paper at ACM SIGPLAN Symposium on Scala, which was held in St. Louis, U.S. on September 28.

Sung-ho Lee, Ph.D. student of KAIST School of Comp..

LeeSeong-ho(Advisor Ryu Seok-young) School of Computing, KAIST was selected as the best reviewer for the Artifact Evaluation at OOPSLA 2018 (Object-Oriented Programming, Systems, Languages, and Applications), a class S international institute in Boston, U.S. on Nov. 7-9 and won the Distinguished Artifact Evaluation Committee Member Award.

APLAS 2018 International Congress was chaired by P..

The 16th APLAS (Asian Symposium on Programming Language and Systems), which KAIST supported as a Gold Sponsor, was held in Wellington, New Zealand, December 3, 2018. Professor Sukyoung Ryu, chair of the program, invited three world-renowned lecturers: Amal Ahmed (USA), Azalea Raad (MPI-SWS, Germany), and Bernard Scholes (University of Sydney, Australia), and conducted various programs including presentations of papers and poster sessions and workshops on December 2 and December 6, 2018.

Professor Jinah Park delivered a keynote lecture a..

Professor Jinah Park of the School of Computing at KAIST delievered a keynote lecture on the topic “Hippocampal Morphology Study based on Progressive Template Deformable Model” at the International Conference on Medical Imaging and Case Reports 2018 held in Baltimore in the U.S. from October 26 to 28, 2018. The event involved 70 people from 48 institutions in 23 countries. The event attracted many researchers who are influential in medical images and clinical studies including Dr. Marth Shenton from Harvard Medical School, Professor Polina Golland of MIT, Michael Miller and Professor Jeffery Siewerdwen of Johns Hopkins University.

Top 15 Most Impactful Early Career Researchers in ..

Professor Shin Yoo has been identified as one of the top 15 most impactful early career researchers in software engineering by a bibliographic study published in the Journal of Systems and Software in October 2018. The result is based on analysis of 14,456 papers as well as their citation counts published in top journals and conferences between 2010 and 2017. Prof. Yoo is an associate professor at the KAIST School of Computing. From 2012 to 2015, he was a lecturer of software engineering in the Centre for Research on Evolution, Search, and Testing (CREST) at the University College London, UK. Prof. Yoo received a Ph.D. in Computer Science from King’s College London, UK, in 2009. His main research interest is in Search Based Software Engineering (SBSE), specifically, the use of meta-heuristics and computational intelligence, such as genetic algorithm, to automatically solve various problems in software engineering, especially those related to testing and debugging.

The most influential paper among those published a..

Together with Cristiano Calcagno, Dino Distefano and Peter O’Hearn at Facebook, Professor Hongseok Yang received the ACM SIGPLAN Most Influential POPL Paper Award for the paper “Compositional Shape Analysis by means of Bi-abduction,” which they presented at POPL 2009. POPL is a top international conference in programming languages. This annual award is given to the authors of the most influential paper among those published at POPL ten years ago. Prof. Yang is a full professor at the School of Computing at KAIST. His research area is programming languages, but he mostly works on problems that arise at the boundaries between programming languages and other areas, such as machine learning, statistics, probability theory and distributed systems. More information can be found in Facebook Research’s article: https://research.fb.com/popl-2019-most-influential-paper-award-for-research-that-led-to-facebook-infer/

Professor Min H. Kim has chosen the best research ..

Professor Min H. Kim has chosen the best research and development of 2018 top 100 Six professors from KAIST have been selected for the "Top 100 2018 state research and development" announced by the Ministry of Science and ICT and the Korea Institute for Science and Technology Planning. The winners are Il-du Kim and Byeong-gook Park at the Department of New Material Engineering, Ho-min Kim at the Graduate School of Medicine, Jae-woo Lee at the Department of Biochemistry, Min H. Kim at the School of Computing, and Kyung-chul ​ Choiat the Department of Electrical and Electronic Engineering. Professor Min H. Kim of the School of Computing was chosen for excellence in the field of information and electronics. It was recognized that he developed high-performance imaging technology to obtain high-performance images. Professor Il-du Kim was chosen as the best performance in the field of machinery and materials. It was recognized that he developed ultra-high sensitivity gas sensor platform material based on self-assembly organic complex catalyst coupling. Professor Byung-kook Park was also chosen for excellence in the machine and material fields. He has developed a material technology that obtains spin current from heat. Ho-min Kim , a professor at the medical school, was selected as the best performance in the life and ocean fields. It has established the 3rd structure and molecular warfare of core protein that controls the formation of synapses. Professor Jae-woo Lee has been chosen for excellence in energy and the environment. He was credited with developing high-value carbon-material synthesis technology. Kyung-chul Choi was chosen as the best performance in information and electronics. It was recognized that it developed a smooth display that was implemented on top of clothes. The performance selected for excellence will be awarded to the Minister of Science and ICT with a certificate and a signboard. The six selected professors will be nominated as candidates for the state-funded R&D achievement evaluation (medal, presidential citation, and the prime minister's citation) and will receive preferential treatment in the selection of new R&D projects.

KAIST School of Computing Professor Eun-Ho Yang a..

"DropMax," "Adaptive Variable Softmax," "Uncertified-Aware Attachment," a joint study by Professor Yang Eun-Ho and Professor Hwang Seong-ju of the KAIST School of Computing, were published by Neural Information Processing System (NIPS) 2018 (acceptance rate:20.8 ％), the world's best academic institution. The paper was published as follows: [DropMax: Adaptive Variational Softmax] In this paper, we propose techniques to improve the performance of the systeresis by applying Dropout techniques to softmax functions that are originally applied to multi-class identification. At each learning quarter, the whole class' subset is randomly sampled according to the learned Dropout probability and the discrimination is performed. There is meaning that is is a ensemble of different discriminations', and it also has the effect of making more comparisons between classes that are likely to be confused with each other. Deep Bayesian techniques enable the learning and reasoning of this probabilistic neural network. Specifically, it has been experimentally verified in this paper that inducing a Dropout mask, the probability of being a target label for each class, differently for each input instance and correctly deduplicating the values is important for improving performance, thereby enabling more reliable learning and reasoning. [Uncertainty-Aware Attention for Reliable Interpretation and Prediction] Existing in-depth neural networks cannot take into account uncertainty in the in-depth learning process of predictive models, so the learning model has the potential to predict the wrong situation, and the uncertainty present in the data and the model itself has led to false predictions. The "Uncertified-Aware Attachment for Reliable Interprevention and Prescription" paper is a way to overcome existing uncertainty problems. The "Uncertified-Aware Attachment for Reliable Interprevention and Prescription" paper is a way to overcome existing Uncertainty problems. Uncertainty-Aware Attachment Mechanism, which allows input-adaptive organic learning by identifying data and model conditions regarding uncertainty, was presented. This paper has experimentally demonstrated that, with respect to the performance of modern machine learning algorithms, the ability to interpret accurate models with corrected model predictive performance and uncertainty is outstanding through UA framework. [Joint Active Feature Acquisition and Classification with Variable-size Set Encoding] This paper addresses the question of learning policies to obtain effective information at a lower cost in situations where the cost of obtaining the necessary information in the classification (e.g., examination costs) is incurred, such as diagnosing diseases. When diagnosing patients, it is reasonable to perform the necessary tests according to the individual's condition rather than carrying out and diagnosing all the same types of tests. In this paper, it jointly learns about dynamic feature acquisition policies and classification models that determine the required information sequentially based on the information currently in place. To this end, we defined a function for the purpose of considering both the classification performance and the costs necessary to obtain information, and presented a joint learning framework for the feature acquisition policy and classification model using in-depth hardening learning. In addition, experiments using synthetic and medical data have shown that the performance is cost-effective compared to static feature acquisition methods or previous studies.

Intelligent Service Integration (ISI) Laboratory o..

Intelligent Service Integration (ISI) Laboratory of KAIST School of Computing(Prof. Han Dongsoo) has decided to carry out the international academic project for the development of Global Indoor Positioning System with Global H Company. Global H will support KAIST ＄ 400,000 for the project, which will be accompanied by technology transfer. The ISI Lab, which developed the KAIST Indoor Positioning System (KAILOS) in 2014 and released to the public, has recently developed a crowd sourcing AI technique that automatically labels the location of wireless signals acquired from unspecified number of smartphone users. In this joint international project, KAIST will upgrade location labeling AI technique using crowdsourcing to commercialization level. If the project is successfully performed, H Company is expected to install the system developed by KAIST its products and commercialize it. Professor Han Dongsoo, who will lead the international industry project with H company to be conducted over the next year, said, "It is a pity that the technology developed by KAIST will be commercialized through overseas companies before the domestic companies. The result is that domestic regulations on location services are too strict. " Professor Han Dongsoo said, "It is significant that a global company recognized the technology developed by KAIST and commissioned an industrial project for commercialization. Technology continues to decline and disappear if not developed. We will continue to expand international industry-academia cooperation in the field of indoor location recognition in the future. "

Professor Kwang-Jo Kim, publication on network int..

Kim Kwang, professor of KAIST (President: Sang Chul Shin) Graduate School of Computing and Information Security Graduate School, along with Muhamad Erza Aminanto and Dr. Harry Chandra Tanuwidjaja, Ph.D. students from Indonesia, wrote an English book called “Network Intrusion Detection using Deep Learning : A Feature Learning Approach” (in the aspected of network intrusion detection using features of deep learning) which supported by IITP’s project on “Communication Technology Research using Bio-inspired Algorithm” from April 2018 to February 2018. He has published this in one of the cyber security systems and networking series at Springer, a prominent publisher in Germany. This book introduces various methods of intrusion detection system using deep learning, which is artificial intelligence technique and widely used in computer vision, natural language processing, image processing, etc., and described an intrusion detection technique which extracts and learns the features of intrusion traffic with very high scan rate (99.918％) and low false positives (0.012％) compared to existing skills. Annexes include technical trends on artificial intelligence techniques for detecting malicious codes. This book will serve as a good guide for undergraduate and graduate students, R & D personnel in providing practical knowledge of establishing cyber security systems regarding graft artificial intelligence onto cyber security. 참조: https://www.springer.com/gp/book/9789811314438

Professor Moonzoo Kim delivered a keynote speech a..

Professor Moonzoo Kim delievered a keynote speech at the 15th International Conference on Formal Aspects of Component Software (FACS) which was held from October 10 to 12, 2018 with Professor Edward Lee of UC Berkeley and Professor Gigore Rosu of UIUC. His talk was titled “Lessons Learned from Automated Analysis of Industrial SW for 15 Years”.

Professor Sung Ju Hwang, selected as the winner in..

Professor Sung Ju Hwang of School of Computing was selected as the winner in the Google AI Intensive Research Awards program announced on September 13th through Google Korea blog. related article: https://korea.googleblog.com/2018/09/googleaifocusedresearchawardsprogram.html