Autonomous Vulnerability Patching Agent — DARPA AIxCC
Role: Patching Component Lead, Team Shellphish (Aug 2023 – Aug 2025)
The DARPA Artificial Intelligence Cyber Challenge (AIxCC) is a global competition for automated vulnerability discovery and repair in open-source and critical-infrastructure software. Team Shellphish was named one of 7 finalists in Aug 2024 ($2M finalist award) and placed 5th in the final round in Aug 2025.
Vulnerability patching agent
- Led development of a multi-agent vulnerability repair system integrating autonomous triage, patch generation, and validation across distributed services.
- Built a feedback loop combining fuzzing, proof-of-vulnerability re-tests, and deterministic regression tests, so that only patches that eliminate the vulnerability and preserve functionality are accepted — >95% patch correctness in the final scored rounds.
- Triaged and patched 30+ C/Java vulnerabilities in projects including curl, NGINX, and Apache ZooKeeper, covering buffer overflows, use-after-free, command injection, and path traversal.
Proof-of-concept generation agent
- Developed a multi-agent system that combines static-analysis results (CodeQL) and code context to generate inputs targeting specific vulnerabilities.
- Built a reachability-based program-slicing pipeline to focus agents on relevant code paths, using coverage-guided search to bypass blocking logic.
- Generated PoCs in ~5 minutes on average, triggering 10 vulnerabilities in ZooKeeper, Tika, and jsoup — five of which Jazzer missed after 8 hours of fuzzing.