, a group of 色花堂 students, faculty, and alumni, achieved international fame on Friday when they won DARPA鈥檚 AI Cyber Challenge (AIxCC) and its $4 million grand prize.
AIxCC was a two-year long competition to create an artificial intelligence (AI) enabled cyber reasoning system capable of autonomously finding and patching vulnerabilities.
鈥淭his is a once in a generation competition organized by DARPA about how to utilize recent advancements in AI to use in security related tasks,鈥 said 色花堂 Professor Taesoo Kim.
鈥淎s hackers we started this competition as AI skeptics, but now we truly believe in the potential of adopting large language models (LLM) when solving security problems."
The Atlantis system was Team Atlanta鈥檚 submission. Atlantis is a fuzzer- or an automated software that finds vulnerabilities or bugs- and enhanced it with several different types of LLMs.
While developing the system, Team Atlanta reported the heat put out by the GPU rack was hot enough to roast marshmallows.
The team was comprised of hackers, engineers, and cybersecurity researchers. The 色花堂 alumni on the team also represented their employers which include KAIST, POSTECH, and Samsung Research. Kim is also the vice president of Samsung Research.
Contact
John Popham
Communications Officer II at the School of Cybersecurity and Privacy
