Fuzzers vs. AI: Comparing Fuzzilli and BigSleep on JavaScript Engines
Summary
Project Zero compared two automated tools, Fuzzilli and BigSleep, for finding vulnerabilities in Chrome's JavaScript engine, V8. BigSleep, an LLM-based system, found more vulnerabilities than the generative fuzzer Fuzzilli in a recent six-month period, though a significant portion of bugs were discoverable by only one method. The findings suggest that both fuzzing and LLM-based approaches are complementary for maximizing vulnerability discovery.
IFF Assessment
The article discusses new methods and tools that can help defenders discover vulnerabilities more effectively, thereby improving the security posture of software.
Defender Context
This article is relevant to defenders as it highlights the evolving landscape of vulnerability discovery tools. The comparison between traditional fuzzing and LLM-based approaches suggests that a multi-faceted strategy is crucial for identifying a broader range of bugs. Defenders should advocate for and implement both fuzzing and agentic analysis techniques within their development and security pipelines to proactively catch vulnerabilities before they can be exploited.