LLMs for Vulnerability Discovery: Finding 0days at Scale With a Click of a Button

No ratings

Presented at Wild West Hackin Fest 2025 by

Marcello Salvati

Dan McInerney

Large Language Models (LLMs) have opened up the floodgates for a whole new generation of security tooling. One of the most obvious applications is automatic discovery of vulnerabilities which so far has had extremely mixed results. Can LLMs “get good” at vulnerability discovery? In this talk, we cover our approach to the problem going into all the success and fails along the way. Finally, we will be tool dropping VulnHuntr, which implements our approach to using LLMs for discovering vulnerabilities through static code analysis along with presenting a number of 0days that were found by it.