ELFuzz: Efficient Input Generation via LLM-driven Synthesis Over Fuzzer Space
Chuyang Chen
34th USENIX Security Symposium (USENIX Security '25) · Day 3 · Software Security 3: Fuzzing
This talk introduces **ELFuzz**, a novel evolutionary approach designed to efficiently generate high-quality seed test cases for mutation-based fuzzing. Presented by Chuyang Chen, a PhD student at Ohio State, the research is a collaborative effort with Professor Brandon Dolan Gabit from New York University and Chen's advisor at Ohio State. ELFuzz addresses a critical challenge in software security: the need for syntactically and semantically valid initial inputs that can effectively guide fuzzers to discover deep program logic and vulnerabilities.
AI review
ELFuzz is legitimate research with a clean core idea: stop using LLMs as dumb input generators and instead use them to evolve the generators themselves, with coverage feedback closing the loop. The fuzzer space lattice — selecting survivors by collective code range coverage rather than raw line counts — is the genuinely clever bit, and the ablation study confirms it carries the system. Five real bugs in a real target keeps this grounded.