"Security Radar: Benchmark Data for Autonomous Vulnerability Discovery"
收藏资源简介：
"Benchmark Data for Autonomous Vulnerability Discovery is a dataset for Security Radar, an autonomous AI agent for grammar-guided vulnerability discovery. Contains 23 scan results across 5 operating systems (Alpine, Ubuntu, Debian, CentOS, Windows 10), 4 web frameworks (Flask, Express.js, Django, FastAPI), and 4 LLM evaluators (Qwen3-32B, Gemma 4 31B\/E4B\/26B-A4B). Includes depth-3 and depth-6 chained attack results, WAF (ModSecurity) robustness tests, sqlmap comparison baselines, and per-vulnerability-type recall analysis.The formal attack grammar covers 20,736 patterns across 8 injection types with admissible pruning achieving 62-100% search space reduction. Summary CSVs reproduce all paper tables. Target application source code and reproduction scripts are included for full experimental replication."
