Large Language Models for Software Security in the AI Cyber Challenge
Author(s)
Lu, Daniel
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The AI Cyber Challenge (AIxCC) is a contest organized by DARPA where competitors secure software at scale using AI systems. We describe a number of steps in the process for tackling this contest, which largely revolve around utilizing large language models to aid static analysis and fuzzing. In particular, we discuss the implementation of a language model plugin system for improving reasoning about code, as well as the application of language models for generating fuzzing harnesses.
Sponsor
Date
2025-07-22
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis