FuzzingBrain V2: A Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction
2026/05/20 by Ze Sheng, Zhicheng Chen, Qingxiao Xu +2 · 10 voices
#cs.CR #cs.SE
paper · pdf
Abstract
Software vulnerabilities pose critical security threats, with nearly 50,000 CVEs reported in 2025. While Large Language Models (LLMs) show promise for automated vulnerability detection, three key challenges remain. First, LLM-generated vulnerability reports suffer from high false positive rates and lack reproducible verification. Second, existing LLM-based approaches use suboptimal granularities for vulnerability localization: function-level analysis overlooks bugs when context becomes extensive, while line-level analysis lacks sufficient context. Third, existing approaches have difficulty reasoning about vulnerabilities with complex cross-function dependencies and triggering conditions. We present FuzzingBrain V2, a multi-agent system that addresses these gaps through four key contributions: (1) fully automated vulnerability analysis built on Google's OSS-Fuzz, ensuring all reported vulnerabilities are fuzzer-reproducible; (2) Suspicious Point, a novel control-flow-based abstraction for precise vulnerability localization at the optimal granularity; (3) logic-driven hierarchical function analysis with dual-layer fuzzing enhancing function coverage under resource constraints; (4) MCP-based static and dynamic analysis tools with context engineering enhancing complex vulnerability reasoning. On the AIxCC 2025 Final Competition C/C++ dataset, FuzzingBrain V2 achieved 90% detection rate (36 of 40 vulnerabilities). In real-world deployment, FuzzingBrain V2 discovered 29 zero-day vulnerabilities across 12 open-source projects, all confirmed and fixed by maintainers, with 2 assigned CVE IDs.
Citations
Discussions
- Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction [hn, 57 points, 9 comments]
- Multi-agent systems finding bugs autonomously – but I'm curious how they handle the false positive noise that usually drowns out real vulns. https://arxiv.org/abs/2605.21779 [bsky, 1 points, 1 comments]
- 📰 A multi-agent Large Language Model system has been developed for automated vulnerability discovery and reproduction, offering enhanced efficiency and accuracy in cybersecurity research. 🔗 https:// [bsky, 1 points, 0 comments]
- Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction https://arxiv.org/abs/2605.21779 (https://news.ycombinator.com/item?id=48297723) [bsky, 1 points, 0 comments]
- Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction https:// arxiv.org/abs/2605.21779 # arxiv # llm [mastodon, 0 points, 0 comments]
- Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction https://arxiv.org/abs/2605.21779 (https://news.ycombinator.com/item?id=48297723) [bsky, 0 points, 0 comments]
- Introducing a cutting-edge multi-agent LLM system designed for automated vulnerability discovery and reproduction. This innovative approach streamlines security testing, enhancing efficiency and accur [bsky, 0 points, 0 comments]
- FuzzingBrain V2: A Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction arxiv.org/abs/2605.21779 [bsky, 0 points, 0 comments]
- Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction https://arxiv.org/abs/2605.21779 [bsky, 0 points, 0 comments]
- Tonight's reading: FuzzingBrain V2: A Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction arxiv.org/abs/2605.21779 [bsky, 0 points, 0 comments]
Related