From Detection to Understanding : Leveraging LLMs in Web Vulnerability Scanners

This study presents a framework that integrates web vulnerability scanners with a Large Lan-guage Model to generate reports tailored to different expertise levels. In our initial experimentsand limited user interviews comparing our developed framework with established tools, weobserved that while our solution required longer processing times, the participants reported im-proved vulnerability comprehension and found the remediation guidance helpful. Most of oursample of users expressed a preference for our developed framework over traditional scanners,mainly citing a better understanding of security issues. These preliminary findings suggest thatLLM-enhanced security tools may help address challenges in translating vulnerability detec-tion into effective remediation; although a more extensive evaluation is needed to confirm theseinitial observations.