WAFPlanet Blog

Insights, tutorials, and news about Web Application Firewalls and application security.

All Posts Best Practices Comparisons Research Security News Tutorials

Autoresearch trajectory showing balanced accuracy improving from 0.630 to 0.976 across 20 experiments

Research · Mar 16, 2026 · 15 min read

An AI Agent Improved OWASP CRS Detection by 80% in 20 Experiments

We pointed an AI agent at the actual OWASP Core Rule Set regex patterns and let it find bypasses, fix them, and reduce false positives. 20 experiments, 20 kept, 0 discarded. TPR went from 55.8% to 100%, FPR dropped from 29.7% to 4.8%.

research modsecurity crs

Experiment trajectory visualization showing balanced accuracy improving from 80.8% to 98.4% across 19 autonomous experiments on CRS v4.24.0

Research · Mar 15, 2026 · 18 min read

We Let an AI Agent Run Autonomous Experiments on ModSecurity CRS. Here's What It Found.

Starting from stock OWASP CRS defaults, an AI agent improved balanced accuracy from 86.7% to 96.7% on CRS v3.3.8, and from 80.8% to 98.4% on CRS v4.24.0, running 30 experiments over 7 hours. Full methodology, results, and downloadable configs.

research modsecurity crs