본문 바로가기 대메뉴 바로가기

KAIST

NEWS

홈페이지 통합검색

-
KOREAN

%72%65%73%65%61%72%63%68%20%73%63%68%6f%6c%61%72%20%61%77%61%72%64

Professor Youngjin Kwon's Team Wins Google Award 'Catches Bugs Without a Real CPU < Professor Youngjin Kwon > Modern CPUs have complex structures, and in the process of handling multiple tasks simultaneously, an order-scrambling error known as a 'concurrency bug' can occur. Although this can lead to security issues, these bugs were extremely difficult to detect using conventional methods. Our university's research team has developed a world-first-level technology to automatically detect these bugs by precisely reproducing the internal operation of the CPU in a virtual environment without needing a physical chip. Through this, they successfully found and fixed 11 new bugs in the latest Linux kernel. Our university announced on the 21st that the research team led by Professor Youngjin Kwon of the School of Computing has won the 'Research Scholar Award' (Systems category) presented by Google. The Google Research Scholar Award is a global research support program, implemented since 2020, to support Early-Career Professors conducting innovative research in various fields such as AI, Systems, Security, and Data Management. It is known as a highly competitive program, with the selection process conducted directly by Google Research scientists, and only a tiny fraction of the hundreds of applicants worldwide are chosen. In particular, this award is recognized as one of the most prestigious industry research support programs globally in the field of AI and Computer Systems, and domestic recipients are rare. ■ Technology Developed to Detect Concurrency Bugs in the Latest Apple M3 and ARM Servers Professor Kwon's team developed a technology that automatically detects concurrency bugs in the latest ARM (a CPU design method that uses less power and is highly efficient) based servers, such as the Apple M3 (Apple's latest-generation computer processor chip). A concurrency bug is an error that occurs when the order of operations gets mixed up while the CPU handles multiple tasks simultaneously. This is a severe security vulnerability that can cause the computer to suddenly freeze or become a pathway for hackers to attack the system. However, these errors were extremely difficult to find with existing testing methods alone. ■ Automatically Detects Bugs by Reproducing CPU Internal Operations Without a Real CPU The core achievement of Professor Kwon's team is the 'technology to reproduce the internal operation of the CPU exactly in a virtual environment without a physical chip.' Using this technology, it is possible to precisely analyze the order in which instructions are executed and where problems occur using only software, without having to disassemble the CPU or use the actual chip. By running the Linux operating system based on this system to automatically detect bugs, the research team discovered 11 new bugs in the latest Linux kernel* and reported them to the developer community, where they were all fixed. *Linux kernel: The core operating system engine that forms the basis of servers, supercomputers, and smartphones (Android) worldwide. It acts as the 'heart' of the system, managing the CPU, memory, and storage devices. Google recognized this technology as 'very important for its own infrastructure' and conferred the Award. < Google Scholar Award Recipient Page > This technology is evaluated to have general applicability, not only to Linux but also to various operating systems such as Android and Windows. The research team has released the software as open-source (GitHub) so that anyone in academia or industry can utilize it. Professor Youngjin Kwon stated, "This award validates the international competitiveness of KAIST's systems research," and "We will continue our research to establish a safe and highly reliable computing environment." ※ Google Scholar Award Recipient Page: https://research.google/programs-and-events/research-scholar-program/recipients/ GitHub (Technology Open-Source): https://github.com/casys-kaist/ozz
2025.11.21 View 1353