A new framework aims to make web application fuzzing more precise by instrumenting the runtime interpreter itself. The system, called ZIMPAF, uses multi-granular interceptors to gather detailed execution traces, including branch coverage and error events. These traces feed into RedPhuzz, a fuzzer that leverages the extra information to generate and guide test cases.

The approach goes beyond simple input mutation by monitoring language constructs and function calls, giving the fuzzer a deeper view of how an application behaves. This high-fidelity monitoring is designed to uncover bugs that might be missed by traditional coverage-guided fuzzing, which often relies on coarse-grained feedback.

The paper describes the architecture of both components and how they work together, but the abstract does not yet include experimental results or comparisons with existing fuzzers. As such, the practical effectiveness of the framework remains to be demonstrated in future work.