KAIST - COMPUTER SCIENCE

  • korea
  • search
  • login

Directions

  • Designing a service robot that can interpret human..

    Reference article: https://zdnet.co.kr/view/?no=20220421180134 "Making a service robot that can interpret human emotions with AI" This is an interview with Professor Sung-Eui Yoon and Professor Sungho Cho from the KAIST School of Computing. ■ A service robot that assists by understanding human behavior KAIST SoC Professor Sung-Eui Yoon and Professor Sungho Jo jointly conducted research on ‘Robot social interaction technology’. The key to this technology is understanding human behavior and assisting humans using the camera on the robot. Existing robots were limited to performing specific tasks, but these robots have the advantage of being able to help in more diverse situations. Professor Yoon explained, “Robots are interesting at first, but the more you use them, the more you discover the shortcomings of their simplicity." “The goal of our research with KT is to address these shortcomings and create a robot that can cope with various environments,” he added. Prof. Sung-Eui Yoon Robots equipped with social interaction technology can predict human behavior to help with tasks. Professor Yoon explained this using a cooking robot as an example. He said, “When cooking, the existing robots could only help when we cook according to a set scenario. However, with social interaction technology, robots can help even if we change the order of operations.” Professor Jo explained that robot social interaction technology can be applied to all service robots. “The technology can be applied to all robots that serve people,” he said. “For now, we are developing a robot according to general situations, and when there is a specific application case, it will be optimized and released as a product.” Prof. Sungho Jo Professors Yoon and Jo consider a robot that can recognize human emotions as the final stage of robot social interaction technology. Professor Yoon said, “Current technology has broadened the set of robots' actions by recognizing human behavior. If research in related fields continues, it is possible to create a service robot that can read people's emotional state and respond accordingly."

    ...Read more
  • KAIST signs an MoU with Busan Metropolitan City to..

    On March 3rd, KAIST decided to establish Auto-ID Labs Korea (Center Director: Professor Daeyoung Kim) in Centum City, Busan, and signed an MoU with Busan Metropolitan City. Auto-ID Labs are an international joint research network for Internet of Things and are composed of premier universities from 6 countries - KAIST, MIT, Cambridge, ETH Zurich/St. Gallen, Fudan, Keio. It is a primary research partner of GS1, a non-profit international organization for developing and maintaining standards. The term Internet of Things was introduced by the Auto-ID Center in 1999 for the first time to the world. http://autoidlabs.org Relevant article : https://busanmbc.co.kr/article/blT21U9obk-N-AQu_

    ...Read more
  • Dae-Young Park, a Master’s student of School of Co..

    Dae-Young Park, a Master’s student (advisor: Prof. In-Young Ko) of School of Computing at KAIST received the ‘Best Paper Award – 1st Place’ for the paper titled, “Urban Event Detection from Spatio-temporal IoT Sensor Data Using Graph-Based Machine Learning” in the 2022 IEEE International Conference on Big Data and Smart Computing (BigComp 2022) held in Daegu, Korea from 17th to 20th of January, 2022. In this paper, to meet the challenge of having less accuracy of detecting urban events as the granularity of processing data in the spatial dimension becomes finer, Mr. Park proposes a novel graph-based approach that analyzes geo-spatial characteristics of urban sensor data over time to keep the accuracy of detecting urban anomaly in finer grained geo-spatial scales. Through experiments using real-world urban datasets, he shows that the proposed approach effectively addresses the challenge and outperforms the popular machine-learning-based urban event detection methods.

    ...Read more
  • 「MICCAI 2025」Holing Medical Image Computing and In..

    MICCAI 2025 General Co-Chairs: Professor Jinah Park, School of Computing, KAIST The 28th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2025), a major international conference in the fields of medical imaging, artificial intelligence, and medical robot intervention, has been confirmed to be hosted in Daejeon, Korea. The MICCAI International Academic Event is the largest and the most prominent international conference in the medical technology field, held annually by continental tour, and is held for the first time in Korea. About 2,000 domestic and foreign experts will give presentations and join the workshops, challenges, tutorials, and exhibitions about the latest research on medical imaging technologies that encompass medical image processing, artificial intelligence, medical robotics, telemedicine, computer-assisted diagnosis, and so on. In particular, a special program will be prepared which KAIST-related laboratories are participating in, in Daejeon, 2025. MICCAI 2025 Co-Chairs are Professor Jin-ah Park of School of Computing, KAIST, professor Polina Golland of College of Engineering, MIT, and professor Jong-hyo Kim of Medical School, Seoul National University. For successful hosting, the Korean Medical Societies Advisory Board is organized cooperating 4 major domestic societies: the Korean Society for Medical Imaging and Information Science, the Korean Society of Medical and Biomedical Engineering, the Korean Medical Robotics Society, and the Korean Society of Radiology. With full support from Daejeon Metropolitan City, about 40 researchers from 11 countries around the world are participating in the organizing committee and they plan to prepare international events under the themes of "From Concept to Clinical" and "Pan-Asian Connection". Related Articles : http://www.daejonilbo.com/news/newsitem.asp?pk_no=1500026 https://www.yna.co.kr/view/AKR20211226020600063?section=news http://www.dailycc.net/news/articleView.html?idxno=669117

    ...Read more
  • Professor Juho Kim, Keynote Lecture at AAAI HCOMP ..

    At the AAAI HCOMP 2021 (The Ninth AAAI Conference on Human Computing and Crowdsourcing), Professor Juho Kim of the School of Computing, KAIST gave a keynote lecture titled "Crowd-Powered Interactive Systems: Lessons and Future Directions." HCOMP is an academic conference hosted by AAAI dealing with researches related to human computing and crowdsourcing that effectively combines human intelligence and machine intelligence at the interface of artificial intelligence and human-computer interaction research. Professor Juho Kim gave a keynote lecture sharing the lessons and implications that he felt from his experience of designing a crowdsourcing-based system in various areas such as education, video viewing, civic participation, and decision-making for more than 10 years and suggested the future steps to be taken. The keynote lecture video is available at https://slideslive.com/38972403/juho-kim-keynote.

    ...Read more
  • Professor Euijin Lee(School of Computing, KAIST) a..

    Professor Ui-Jin Lee of School of Computing, KAIST has been appointed as the Program Chair for the ACM UbiComp (International Joint Conference on Pervasive and Ubiquitous Computing), which will be held online from September 21-26, and will be in charge of the organization of the academic program. The ACM UbiComp conference is the international academic conference of the highest authority sharing researches on the design, development, demonstration, and understanding of ubiquitous computing systems for next-generation computing environments including mobile, wearables, Internet of Things (IoT), and virtual/augmented reality (AR/VR). This year, the 30th anniversary of the declaration of the vision of “Ubiquitous Computing”, in commemoration of Dr. Mark Weiser, a pioneer in the field, major scholars such as Professor Gregory Abowd of Northwestern University will discuss “the direction of future research”. Various academic exchange programs such as presentation of more than 230 papers, 38 posters/demos, 11 workshops, 4 tutorials, and doctoral symposiums will be held online. Professor Euijin Lee has published the largest number of papers as a Korean at the ACM UbiComp conference and was appointed as the chair of the program in 2021 based on his contribution to the ubiquitous computing field for a long time. UbiComp Conference website: https://www.ubicomp.org/ubicomp2021/

    ...Read more
  • Online workshop on ˝Computational Knot Theory˝

    Computational Knot Theory is an interdisciplinary research field on the intersection between Computer Science and Pure Mathematics. It has received global attention due to Lisa Piccirillo’s 2020 breakthrough regarding Conway's 50 year old knot problem. Alexander Stoimenov is a renowned expert on Computational Knot Theory, and has recently been able to further his research using extensive CPU time from KAIST’s compute servers. From May 26 to June 2 he has been hosting the world’s leading scholars on Computational Knot Theory, at a virtual workshop supported by KAIST’s School of Computing: https://stoimenov.net/stoimeno/homepage/ckt/ Eleven talks were given by speakers from KAIST, Korea, Japan, Russia, Europe, and USA; audience similarly covered the world, and ranged from undergraduate students to emeritus professors. This important event has thus established KAIST’s School of Computing among the world’s leading centers in Computational Knot Theory.

    ...Read more
  • Prof. Sukyoung Ryu’s research team won ACM Disting..

    A research team led by Prof. Sukyoung Ryu in our school received ACM SIGSOFT Distinguished Paper Award in ACM/IEEE ICSE 2021 (The 43rd ACM/IEEE International Conference on Software Engineering), which was held on May 23 to 29, 2021. The work was done by graduate students Jihyeok Park, Seungmin An, Dongjun Youn, and Gyeongwon Kim, and Prof. Ryu and it is titled “JEST: N+1-version Differential Testing of Both JavaScript Engines and Specification.” The team received the same award from ACM in 2020, and this gives the team the honor of winning the award for two consecutive years. The International Conference on Software Engineering is a flagship conference in the field of software engineering hosted by ACM/IEEE, and is the top conference designated by the Korean Information Science Society. Detail information can be found in the following web page: https://conf.researchr.org/home/icse-2021

    ...Read more
  • Editorial board member of IEEE Transactions on Sof..

    Prof. Sukyoung Ryu, a faculty member of the School of Computing, KAIST, has been appointed as an editorial board member of IEEE Transactions on Software Engineering (TSE) starting from April 2021. TSE is the top journal in software engineering, and one of two SCI journals in software engineering, along with ACM Transactions on Software Engineering and Methodology (TOSEM). TSE has an impact factor of 6.11 in JCR 2019 and is currently ranked at the 6th place by Google Scholar among the Software Systems conferences and journals.

    ...Read more
  • Kick off AI Joint Research Center, KAIST-Samsung S..

    We signed a memorandum of understanding (MOU) with Samsung SDS (CEO Hwang Seong-woo) to establish and operate the KAIST-Samsung SDS AI Joint Research Center on the 23rd and held a signboard ceremony for the institute. With this MOU, the two organizations cooperate for necessary matters and support projects to make their common goal come true such as ▲to establish and operate Joint Research Center for the research and development of advanced technology in AI field ▲to recommend and recruit talented human resources in the field of AI ▲to secure other AI technologies. At the School of Computing, as artificial intelligence is originated from computer science, about 50 professors are conducting various AI-related research and giving lectures under the vision of human-centered computing. In the future, KAIST-Samsung SDS AI Joint Research Center is going to innovate artificial intelligence technology by creating synergy with essential fields of computer science such as software engineering and human-computer interaction with a focus on natural language processing and knowledge graph technology. Samsung SDS, founded in 1985, is a company in charge of Samsung's ICT and is currently leading the innovation of digital transformation in various industries in over 30 countries. Based on its core IT technologies such as AI, cloud, and blockchain, it provides the foundation for companies to make the most innovative leap in various business areas such as manufacturing, finance, logistics, and retail. In SDS research center, about 300 outstanding domestic and foreign researchers are working hard on studying the technologies based on AI, data analysis, blockchain, cloud, security, and quantum. At the agreement ceremony held on the 23rd, Seong-hyun Maeng, professor of School of Computing, and Lee Sang-wook, director of Samsung SDS Research Center, signed on behalf of both organizations and held a signboard ceremony. Then, at the opening ceremony, Dong-man Lee, dean of the College of Engineering, and Seok-young Ryu, dean of School of Computing, attended to deliver congratulatory speeches and expectations on the development and role of the AIJjoint Research Center. In order to prevent the spread of COVID-19, 10 professors and related professors attended online, while researchers from Samsung SDS, including Sang-wook Lee(director of Samsung SDS Research Center),Young-joon Kwon(head of AI Research Center), Tae-hee Lee(head of ML research team), Tae-hyun Kim(head of Technical strategy team), and Hyo-jin Jeon(group leader) participated in the ceremony on site. Sang-wook Lee, director of Samsung SDS's Research Center, said, "By establishing a mid- to long-term cooperation system with KAIST, we will not only preemptively secure rapidly developing AI technology, but also keep applying and spreading AI technology to lead the industry.” he said. Sung-hyun Maeng, director of KAIST Joint Research Center, said, "The two leading institutions in academia and industry will work together to create a new turning point in developing artificial intelligence technologies domestically and fostering talented human resources in the industry." Additionally, Maeng, director of the Center, emphasized, “It is the purpose and current role of KAIST to support researches in the industry by fostering high-quality scientific and technological talents with in-depth theoretical knowledge and practical application skills in the science and technology field.” We will do our best to make the AI ​​Research Center a good example for industry-academy cooperation that represents the era of the 4th industrial revolution.” [Link]

    ...Read more
  • Technology transfer of Customized Tourism Content ..

    KAIST Professor Dongman Lee's research team at our School proceeded a technology transfer with O2O Co., a software developer and supplier company, on "Customized Tourism Content Recommendation System and Method based on Unstructured Data Crawling, which worths 50 million won (around $50000). This technology has been recognized for its excellence and exclusivity for its use of a decision tree-based popular place extraction technology for each travel theme and a travel scheduling technology based on a multi-criteria shortest distance algorithm. Existing technologies that provide personalized location-based travel content include TripAdvisor, Google Trips, and My Real Trip. They provide information on popular tourist destinations or hotels around the searched area, as well as visitors' evaluation scores and opinions, sorted by popularity or tourist destination type. In addition, they help users to create/edit a personal travel itinerary or recommend a pre-made daily travel itinerary. In these existing technology, a user can create a travel itinerary based on individual tourist destination information or refer to a predefined travel itinerary. However, the new technology can create the best travel itinerary in real-time using the user's desired travel purpose and scheduling priority (minimum cost, shortest distance, etc.). In other words, it does not just recommend individual places based on social big data, but it differentiates from other services by implementing “curation” to reprocess the data in the form of a travel itinerary. This technology can be applied to the online-to-offline (O2O) market, which consumes offline services based on online information. As the scale of the market continuously expands, the market value of such technology is expected to boost up in the future.

    ...Read more
  • Professor Min Suk Kang's research team was nominat..

    The security vulnerability of the 4G/5G mobile communication system discovered by Professor Min Suk Kang's research team at our School was officially acknowledged by the GSMA (World Mobile Communication Operators Association) vulnerability disclosure program (CVD) (CVD-2020-0040). As a result, Professor Kang was named in the Mobile Security Hall of Fame. (GSMA Homepage: https://www.gsma.com/security/gsma-mobile-security-hall-of-fame/) The GSMA (World Mobile Communication Business Association), formed in 1987, is a group of mobile communication service providers and mobile phone manufacturers from around the world. It is an association with more than 700 mobile communication service providers and 200 equipment and mobile phone manufacturers. This is the first time for a research institution in Korea to be recognized by the GSMA security vulnerability disclosure program. The disclosed security vulnerability, first discovered by Professor Kang's research team, is a system side-channel defect that unintentionally exposes information related to real-time mobile data usage of 4G/5G users. The vulnerability requires security attention as it enables an attack that estimates the indoor/outdoor movement path of a mobile data user with high accuracy. The mobile user's path tracking attack using the vulnerability is going to be presented in August next year at Usenix Security 2021, one of the most authoritative computer security conferences. The paper is available through the following link (usenix.org/conference/usenixsecurity21/presentation/lakshmanan). The research was conducted in collaboration with Professor Kang's advising student Nitya Lakshmanan (4th year Ph.D. student at the National University of Singapore) and Professor Mun Choon Chan and Professor Jun Han at the National University of Singapore.

    ...Read more
  • Prof. Heung Kyu Lee, developed commercial image fo..

    Digital Innotech Inc., a KAIST laboratory founded Startup company, and Professor Heung Kyu Lee jointly developed and announced KAICATCH 2.0, commercial software for image forgery detection. Existing technologies are developed in response to specific image forgery and manipulation types and are widely used for academic papers and patents with excellent performance of 95% or higher detection rate. However, when an arbitrary photo is given in an actual distribution environment, the detection rate drops down to 5-10%. For this reason, the practicability of the technology has been constantly questioned. Following Farid's research team in the US in 2015, as the second, they provided a public video analysis service for several years and collected over 300,000 images, which were analyzed to identify the problems of the existing techniques. KaiCatch (an abbreviation of KAIST Catch, meaning forgery caught by KAIST) is the first commercial software for detecting image forgery in Korea, and second worldwide after the United States. It has a very high reliability of over 70-80% for identifying the forgery type and its position even when arbitrary photos are given without any specified forgery type information. The development and release of KaiCatch were reported by KBS, Yonhap News National Broadcasting, Chosun, JoongAng, Maeil Economic Daily, Seoul Economic Daily, Electronic Newspaper, Digital Times, Hankyoreh Newspaper, Kyunghyang Newspaper, and other central media. See the reference articles below. Video (KBS national broadcast, Korean): http://mn.kbs.co.kr/mobile/news/view.do?ncd=5040027 Online article (Daily Economy, Korean): https://www.mk.co.kr/news/it/view/2020/11/1127082/

    ...Read more
  • NHK pays attention to infectious disease tracking ..

    On October 27, 2020, The Chosun Ilbo, a Korean daily newspaper, published an article on the infectious disease tracking technology developed by Professor Dong-Soo Han's research team in KAIST School of Computing. The technology was also introduced in a recent documentary produced by NHK, Japan's national broadcast. < Professor Dong-Soo Han is discussing the working principle of the COVID-19 management system based on a smartphone black box technology developed by his research team (Source: NHK broadcast) > NHK introduced the smartphone black box technology developed by Professor Han's research team as one of the few epidemiological investigation techniques with minimal privacy invasion in the documentary "Protect Privacy or Prevent Spreading?" which highlights COVID-19 quarantine in Korea. This documentary exhibits excellent cases of quarantine in Korea, while also dealing with the problem of privacy invasion that occurs during quarantine investigations such as the Itaewon case. In particular, bringing up the need for epidemiological investigation technology that can protect personal information, it compares and evaluates the technology developed by Prof. Han's team along with Google and Apple's Bluetooth technology. The smartphone black box technology developed by Professor Han's team is characterized by encrypted storage of signal information collected by smartphones such as GPS, Wi-Fi, Bluetooth signals, and sensing data. Just as the black box of an airplane is disclosed only in the event of an accident, the trail information recorded in the smartphone black box is analyzed only when the user is found to be a confirmed case. The diagram above is an overview of the infectious disease management system based on smartphone black box technology. The similarity between the smartphone black box information of a confirmed case and that of another individual is calculated as a score. The research team also developed an algorithm that can compute this similarity score in real-time. The quarantine authorities can judge that the users with high scores are more likely to have had contact with a confirmed case and make appropriate measures to prevent further infection. Professor Dong-Soo Han said, "Although the whole world is paying attention to the success of quarantine in Korea, the citizens are still reluctant to disclose personal information to be tracked by the government". "This infectious disease management system based on smartphone black box technology lets us catch two birds with one stone, which are privacy protection and infectious diseases tracking," he added.

    ...Read more
  • [AI Sapiens Era] ˝AI Feels˝ AEI can understand and..

    The interaction between humans and artificial intelligence (AI) is improving day by day. It has become possible to provide users with various customized conveniences. Today, AI can communicate interactively beyond simply presenting the desired value or analysis result. Since then, people have been hoping for an AI capable of "emotional interaction". The implementation of "Artificial Emotional Intelligence (AEI)" that understands and responds to users' emotions is emerging as a new goal. Of course, implanting a real human-like sensibility in AI is something that can't be promised even in the distant future. However, it is relatively easy to build an AI that understands and responds in its own way. "Affective Computing (AC)" introduced in the mid-1990s was the first conduct of this idea as an academic research topic. The main research direction was to analyze physiological signals such as heart rate and brain activity, as well as facial expressions, postures, and movements, to identify the emotional state of the subject. Later, with the development of machine learning and deep learning technology, it was possible to interpret emotions with more information, and this is how AEI technology was born. The first field is to use advanced recognition functions to develop tools and environments that work in response to users' emotions. The key here is to teach AI to recognize and respond to user emotions without exchanging any explicit commands. KAIST CS Professors Sungho Jo and Byunghyung Kim, are actively working in this field. The research team already developed a technology that detects electroencephalogram (EEG) and heart rate with a device to distinguish 12 emotional states of a person. Research is also underway to enhance various tools and environmental actions by applying "emotional feedbacks". When a user feels unhappy about AI's behavior, Ai will learn it and avoid it in the future. Prof. Sungho Jo said, “If you add emotion to AI, you can go beyond the traditional relationship of simply sending and receiving commands, and become a 'mechanical individual' as if you and AI are one entity performing a task". "AI can become much more fine-grained and advanced with emotional feedback," he explained. Research on "coating" human emotions to AI products is also active. Even with a naive imitation, smoother communication will be possible if AI can respond to user emotions immediately and closely. An example is "Ever" being developed by the Korea Institute of Industrial Technology. Ever is a robot that expresses detailed emotions with facial expressions. In the case of 'Ever-6' implemented at the end of last year, it is possible to make more than 12 facial expressions using 26 motors (actuators). The facial expressions of Ever and the researcher can be identified by an emotion recognizer. Facial expressions are the result of emotional interactions. Ever can simultaneously recognize six emotions from multiple faces at various angles and show appropriate facial expression. For example, when people laugh, Ever laughs along. Dr. Dongwook Lee, a researcher at Korea Institue of Industrial Technology, said, “Interaction with a robot induces a sense of repulsion because it doesn't seem like it is actually communicating with us". "Allowing a robot to express emotions through natural dialogues, facial expressions, gestures, and eye contacts is the most direct form of AEI," he added. AEI is applied in industry. It is used not only for robots that mimic emotions but also for the emotional driving where the driver interacts with the car, and also for the customer support service industry to recommend content to users by analyzing their emotions. "AEI will grow even bigger in the future, and if AEI technology is effectively integrated into various industry fields, it will become the leading technology," said Prof. Sunhee Kim of the Department of Brain & Cognitive Engineering at Korea University. The table below shows the AEI technology in major fields. (source: Prof. Sunhee Kim at Korea University) Source: etnews (https://www.etnews.com/20200630000148) Date of Issue: Jun 30th, 2020

    ...Read more
  • Professor Juho Kim appointed as Paper Chair for AC..

    KAIST CS Professor Juho Kim has been elected as Paper Chair for the 23rd ACM Computer-Supported Cooperative Work and Social Computing (CSCW) Conference. ACM CSCW is one of the representative conferences hosted by SIGCHI, ACM's human-computer interaction (HCI) expert group, and is a global conference that ranks second in the h5-index criteria of Google Scholar HCI field. Prof. Juho Kim is the first Paper Chair from an Asian institution in the history of the academic conference that began in 1986. He has been working as a co-chairman with Prof. Sarita Schoenebeck of the University of Michigan and Doctor Siân Lindley‬ of Microsoft Research since last year. For the first time in CSCW's history, he introduced a quarterly thesis deadline and is preparing for the online conference to be held in October. ‬ Prof. Kim received the ACM Recognition of Excellent Service Award in recognition of his contributions as a Paper Chair. ACM CSCW 2020 Website : http://cscw.acm.org/2020

    ...Read more
  • Prof. Sukyoung Ryu’s research team won ACM SIGSOFT..

    A research team led by Prof. Sukyoung Ryu in our school will receive ACM SIGSOFT Distinguished Paper Award in IEEE/ACM ASE 2020 (The 35th IEEE/ACM International Conference on Automated Software Engineering), which will be held on September 21 to 25, 2020. The work was done by Dr. Sungho Lee, undergraduate student Hyogun Lee, and Prof. Ryu and it is titled “Broadening Horizons of Multilingual Static Analysis: Semantic Summary Extraction from C Code for JNI Program Analysis.” Besides, Dr. Lee was appointed as an assistant professor at Chungnam University from September 2020. The ASE Society is a flagship conference in the field of software engineering hosted by IEEE/ACM, and is the top conference designated by the Korean Information Science Society. Detail information can be found in the following web page: https://conf.researchr.org/series/ase

    ...Read more
  • [COVID-19 Emergency Diagnosis] World's First “COVI..

    This broadcast is about the world's first “Coronavirus Diffusion Prevention System” by Professor Han in our department.

    ...Read more
  • [AI Sapiens era] AI shares emotion with humans

    In the movie 'Her', the male protagonist falls in love with the artificial intelligence (AI) operating system (OS) Samantha with a female voice. When the male returns after his failure, Samantha tells a story and music to release his mood. Samantha is able to comfort the main character because she can read various emotions in the tone of the voice and the words in the conversation. AI’s ability to understand the emotion hasn’t been advanced as dramatically as the movie, however, there are various attempts on such technological developments. IBM showed that AI distinguishes laugh in conversations at the 2019 International Conference on Electrical and Electronics Engineers Signal Processing International Conference. For AI to understand emotions, mass data analysis is important. AI analyzes the patterns in data and grasps the meaning. IBM focused on identifying the class of laugh in the conversations and the meaning behind it. In Korea, AI technology has been developed to identify emotions based on biosignals and behaviors detected from the human body. A research team led by Professor Sungho Jo in our school and Professor Byung-hyung Kim developed an emotion identification system by analyzing biosignals with deep learning technology and also made an algorithm to express the relationship between human emotions and causality by linking emotions and behaviors. The deep-learning-based emotion identification system analyzes the electroencephalogram and heart rate signals generated in the frontal lobe of the brain. It distinguishes 12 emotions such as happiness, excitement, joy, sadness, and boredom. This emotional state was associated with actual user behavior. They analyzed how users’ behaviors affect their emotions. The research team predicts that the algorithm will be used mainly in the field of healthcare such as psychotherapy and behavioral development correction.

    ...Read more
  • Prof. Moonzoo Kim delivered a keynote lecture at A..

    Prof. Moonzoo Kim delivered a keynote lecture, together with professor Darko Marinov from UIUC, at ACM/IEEE International Conference on Automation of Software Test (AST) 2020 from July 15 to 16. Keynote topic: CROWN 2.0: Automated Test Generation for Industrial Embedded Software - 17 Years Journey from Research To Product https://conf.researchr.org/home/icse-2020/ast-2020#Keynotes 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. In the Keynote, Prof. Kim talked about his 17 years journey from research to product. He has conducted leading research in the field of embedded SW testing and automation for the past 17 years and he builds a startup V+Lab (https://vpluslab.kr) to maximize the ripple effects of the research achievements.

    ...Read more
  • Professor Alice Haeyun Oh to Join GPAI Expert Grou..

    Professor Alice Haeyun Oh will participate in the Global Partnership on Artificial Intelligence (GPAI), an international and multi-stakeholder initiative hosted by the OECD to guide the responsible development and use of AI. In collaboration with partners and international organizations, GPAI will bring together leading experts from industry, civil society, government, and academia. The Korean Ministry of Science and ICT (MSIT) officially announced that South Korea will take part in GPAI as one of the 15 founding members that include Canada, France, Japan, and the United States. Professor Oh has been appointed as a new member of the Responsible AI Committee, one of the four committees that GPAI established along with the Data Governance Committee, Future of Work Committee, and Innovation and Commercialization Committee.

    ...Read more
  • KAIST · SNU developed hand gesture measurement tec..

    Technology has been developed to measure the movement of the entire hand with only one sensor patch attached to the wrist A 'skin type sensor' that can accurately measure human motion with minimal data has been developed by combining deep learning technology. It is done by a research team, led by Professor Sungho Jo, in our school and Professor Seung-Hwan Ko in the department of mechanical engineering, Seoul National University (SNU). In order to realize augmented reality (AR), a technology for measuring the movement of the human body is required. Currently, the most widely used measurement method is to use a motion capture camera. However, there is a limitation that measurements can only be possible in the space where the cameras are installed. In order to overcome this, wearable device-based techniques for measuring body movement without spatial restriction have emerged. However, there was the inconvenience of attaching sensors to every hundred of joints and muscles where movement occurs. Thus, the team developed a technique that can detect the movement of the entire hand with only one sensor patch attached to the wrist: they used silver nanoparticles to be covered on the sensor substrate and make cracks occur in the silver nanoparticles when there is movement. By detecting the electrical signal generated by this crack, it is possible to measure minute movements occurring in distant parts. The team also used deep learning technology to allow the model to learn various finger joint movements with minimal data and time spent. Professor Jo said, "It is meaningful in that this technology is able to obtain real-time information more effectively in a real environment from people by using deep learning technology. If this is applied for general usage, the era of universalization of wearable AR technology will come sooner. " The results were published in the international journal Nature Communications. (https://biz.chosun.com/site/data/html_dir/2020/05/20/2020052001750.html)

    ...Read more
  • A research team led by Prof. Jooyoung Lee publishe..

    Prof. Lee Ju-young, in Graduate School of Information Security Research, Eurocrypt 2020 A paper is accepted to Eurocrypt 2020, one of the top conferences in the field of cryptography. The paperwork is titled as “Tight Security Bounds for Double-block Hash-then-Sum MACs” and it is led by a research team of Prof. Jooyoung Lee of the KAIST School of Computer Science (Graduate School of Information Security) Message authentication codes (MACs) that update the internal state of 2n-bits using n-bit block ciphers have been proven to be safe up to 2n/3 bits. Recently, after 3n/4 bit-level attacks have been proposed, an exact analysis of safety was considered an important open problem. This paper developed the mirror theory (the theory that approximates the number of equations and inequalities) and proved the safety of MAC up to 3n/4 bit level, and applied it to various MACs such as PolyMAC, SUM-ECBC, PMAC_Plus, 3kf9, LightMAC_Plus, etc. This paper is the only domestic paper for Eurocrypt 2020, and it was awarded the Grand Prize at the National Crypto Contest last year.

    ...Read more
  • Professor Shin Yoo announced the results of teleco..

    As COVID-19 forces workers across the globe into unplanned remote work, an international team including Prof. Shin Yoo from KAIST revealed that software professionals are experiencing diminished emotional wellbeing and productivity. Launched in twelve languages and completed by more than 2,000 individuals in over 50 countries, the study identified how current work environments and situations are affecting software professionals across industries. Other key findings reveal that poor disaster preparedness, fear related to the pandemic and poor home office ergonomics are exacerbating professionals' reduction in wellbeing or productivity; and that women, parents, and people with disabilities may be disproportionately affected. The team, which submitted the findings today (May 4) to Empirical Software Engineering and arXiv, will use results to make practical recommendations to the software industry when managing employees working from home during social isolation now, and potentially in the future, to improve employee’s productivity and wellbeing.

    ...Read more
  • KAIST and Microsoft developed a realtime high-qual..

    A research team led by Professor Min H. Kim in our department together with MSRA developed a texture fusion technology that enables real-time high-quality 3D video content generation. If technology is commercialized, the ripple effect is expected to be enormous. An RGBD camera (Kinect) is used for the technology and the developed algorithm is used to reconstruct geometry and texture information of a real scene in real-time and even able to accomplish texture mapping that applies the texture to a 3D model in real-time. Only tens of milliseconds are needed to complete the algorithm. Depending on the quality of the camera, 4K resolution or higher results can be produced without time delay. Previously, geometry and texture information could be generated, but it was impossible to generate two pieces of information simultaneously. In particular, real-time texture generation technology did not exist. It took tens of minutes to calculate mapping functions for high-quality texture creation and texture mapping. It was also necessary to generate geometry information before this process. The researchers have created a transform function (warp matrix) that minimizes the gap between the frames of geometry and texture information obtained with an RGBD camera, Kinect. They succeeded in optimizing the location correlation of two pieces of information and improved the quality of the results. They published the result in CVPR 2020, one of the top computer vision conferences. (reference: https://www.etnews.com/20200505000038)

    ...Read more
  • Real-time video transmission security technology

    Technology provides the highest level of video security performance while saving up to 50% of CPU and battery usage A research team led by Professor Myung-Chul Kim in our department developed a real-time video encryption technology for video transmission and a technique that consumes low computational resources (CPU, battery, etc.) used in video transmission equipment such as webcams, video drones, CCTV, augmented reality (AR), virtual reality (VR), etc.

    ...Read more
  • Prof. Insik Shin's research team published a paper..

    A research team led by Prof. Insik Shin introduced their developed technology HFL, which automatically detects bugs in operating systems, in Network and Distributed System Security (NDSS), one of the top security conferences. The research team also includes members of Purdue University, NEC laboratories America, Oregon university, and Seoul National University. (Acceptance Rate: 17.4%). The team developed a technique to identify indirect control flow by a function pointer, the dependency between system calls, and overlapping structure of system call arguments, which are characteristic of the operating system. They also developed a new hybrid fuzzing tool, called HFL, by combining an existing fuzzing tool, Syzkaller, and a symbolic execution tool, S2E. Through this, they could solve the performance degrading problems in the existing hybrid fuzzing and the developed tool enabled to utilize the advantages of the hybrid fuzzing in the operating system. By adapting HFL to Linux, an open-source OS, the research team uncovered 24 new security vulnerabilities and contributed to the stability of Linux by reporting and removing the vulnerabilities. The findings were published in a paper at NDSS 2020, one of the most prestigious conferences in computer security, and the HFL tool will soon be available on GitHub as well.

    ...Read more
  • Prof. Sooel Son's research team published a paper ..

    A research team led by Prof. Sooel Son published a paper, titled FUSE: Finding File Upload Bugs via Penetration Testing in Network and Distributed System Security (NDSS), one of the top security conferences in San Diego, USA from Feb. 24 to 27, 2020. File uploading feature in modern web applications is important in terms of sharing user-generated contents. However, there is a potential security threat in that the file upload function may upload arbitrary files created by attackers to the target server and used to take over the target system. In this study, the team of Web Security & Privacy Lab created a simulation testing tool called FUSE, found 30 new vulnerabilities in file uploading features from 33 popular web applications such as WordPress and Joomla, and received 15 CVEs. In addition, the research team released a mock attack testing tool FUSE on GitHub as well. Paper site: https://wsp-lab.github.io/papers/lee-fuse-ndss20.pdf Code site: https://github.com/WSP-LAB/FUSE

    ...Read more
1 2 3 4 5 6 7 8