PenHeal: A Two-Stage LLM Framework for Automated Pentesting and Optimal Remediation

Abstract

Recent advances in Large Language Models (LLMs) have shown significant potential in enhancing cybersecurity defenses against sophisticated threats. LLM-based penetration testing is an essential step in automating system security evaluations by identifying vulnerabilities. Remediation, the subsequent crucial step, addresses these discovered vulnerabilities. Since details about vulnerabilities, exploitation methods, and software versions offer crucial insights into system weaknesses, integrating penetration testing with vulnerability remediation into a cohesive system has become both intuitive and necessary.

Topics & Concepts