Lightning Talk: AI Cyber Challenge
summary
Qualification round runs five challenge problems: Linux kernel, engine x, SQL light, Jenkins, and Tika, each CRS getting four hours and $100 of LLM credit to find vulnerabilities, pinpoint the introducing commit among 100–200 commits per target, and produce nonfunctional-breaking patches. A Kubernetes-backed orchestration pipeline runs locally or on a three‑node cluster. The system pairs concrete analyses and fuzzing with LLMs; fuzzers find bugs, LLMs excel at generating patches, so models help but do not replace programmatic analysis.
AI-written summary · about
related talks
01h 02mPanel Discussion:The Present and Future of AI and Security
44mAI Showdown: The Good, The Glitchy, and The Risky
44mArtificial Intelligence Cyber Challenge (AIxCC): Overview and Releasing Research as Open Source Software
22mEnabling Security Defenders with AI
44mTL;DR: Applying AI to Security
17mLightning Talk: Navigating the Intersection: AI’s Role in Shaping the Secure Open Sour...
07mLightning Talk: AI Penetration Testing
15mOSS-CRS: Liberating AIxCC Cyber Reasoning Systems for Real-World Open-Source Security