Fuzzing is one of the most effective ways to find elusive software vulnerabilities. Despite years of research, general purpose fuzzers such as AFL++ and Libfuzzer struggle to mutate complex data structures effectively, preventing them from exploring deep functionality. Grammar fuzzing, an alternative fuzzing strategy is much more effective but complex to setup and run. Autarkie abstracts away all complexity and and surpasses all other grammar fuzzers in performance while offering novel features.
Gone are the days of finding bugs in the parser or on the surface of applications. The bugs now lie in the core application logic, well beyond the parser. Fuzzing complex targets such as interpreters, databases or network protocols has always been difficult due to their strict input structures. Autarkie was born out of the need of fuzzing complex and evolving data structures with the convenience of fuzzers such as AFL++. Autarkie leverages a simple insight: the target needs to parse the input, so it must define the structure internally. Macros could be used to gain insight into the structure and build a grammar fuzzer.
Autarkie does not just out perform all other grammar fuzzers, but also offers novel features such as removing grammar derivation and maintenance, ability to learn from other fuzzers, constraint solving and resumable fuzzing campaigns.
Join me for this talk where I go through Autarkie's internals, features, and its development journey.
I will also talk about my journey hunting for bugs with Autarkie and hopefully convince you to use it on yours.
Licensed to the public under https://creativecommons.org/licenses/by/4.0/
about this event: https://program.why2025.org/why2025/talk/BNTTXU/
