Fuzz testing and symbolic test generation both face their own challenges. While symbolic testing has scalability issues, fuzzing cannot uncover faults which require carefully engineered inputs. In this paper I propose a combination of both approaches, compensating weaknesses of each approach with the strength of the other approach. I present my plans for evaluation, which include applications of the hybrid tool to programs which neither of the approaches can handle on its own.
39th International Symposium on Software Engineering