ML4Fz is a ML-assisted extensible Windows DLL fuzzer designed to identify potential stack-based buffer overflow vulnerabilities in functions in Export Address Table (EAT) of the target DLL
Enumerate exported DLL functions and build a clear target surface before fuzzing begins.
Four entry points from project context to the first fuzzing run.
The graph shows target discovery, seed generation, ML-guided mutation, execution, triage and feedback as one fuzzing loop.
Inspect the EAT function list and identify user-configured fuzzing targets.
Automatically generate seed inputs for the target function or use user-specified seeds.
Use ML to adapt the fuzzing strategy and mutate fuzzing data for the selected target.
Provide the fuzzing data to the target function and execute it.
Identify and analyze crash data and branches reached during execution.
Use observed results to provide ML-based feedback for fuzzing strategy and data mutation.