PPhying中文
Curated security research, vulnerabilities, advisories and tools for practitioners.

Cyber Reasoning System: The Origin of AI Security and Agentic Defense

hackerone.com | blog | #ai-security | #agentic-ai | #fuzzing | #history | #hackerone | #darpa | #cyber-reasoning-system | #automated-patching | #ai-security-history
📄 Raw .md⬇️ Download .md🔗 Original ↗

Summary

From Shakey the Robot and IBM Autonomic Computing to DARPA's Cyber Reasoning System, this piece traces how autonomous discover-exploit-patch machines became the origin of today's agentic AI defense.
Published
Collected

Related coverage

blog · blog.trailofbits.com
Our thoughts on AIxCC’s competition format
Trail of Bits dissects DARPA's AIxCC format: automated systems must find and patch vulnerabilities in critical software, with concerns flagged on challenge construction, PoU granularity, and scoring.
blog · hackerone.com
Measuring The Fix: How We Evaluate Frontier Models on Remediation Quality
HackerOne tested frontier models on security fix quality and found the judge mattered more than the model. See how CTEM remediation earns engineer trust.
blog · blog.trailofbits.com
DARPA’s AI Cyber Challenge: We’re In!
Trail of Bits announces it will compete in DARPA's AI Cyber Challenge (AIxCC), fielding an AI/ML-driven Cyber Reasoning System built by its software security and AI experts.
blog · blog.trailofbits.com
How We Fared in the Cyber Grand Challenge
Trail of Bits' Cyber Reasoning System found bugs in 65 of 131 CGC qualifier programs and rewrote 94, but a fatal patching-performance flaw dropped it to 9th place, missing the finals cutoff.
research · theori.io
Building Effective LLM Agents | AI Cyber Challenge
Lessons from building LLM agents for DARPA’s AI Cyber Challenge: decomposing tasks, curating toolsets and adapting to models, applied to vulnerability detection, PoV generation and patching.
research · theori.io
Winning the AIxCC Qualification Round
Theori’s RoboDuck cyber reasoning system took first place in DARPA’s AIxCC qualification round, earning the most achievements and unique bug classes and securing $2M plus a 2025 finals spot.
Back