Distributed evolutionary fuzzing with evofuzz

Abstract

This paper describes the design of a tool (called Evofuzz) that implements the technique of evolutionary (or coverage-guided) fuzzing in a scalable, distributed manner. The architecture, design-choices and implementation specifics of this tool are examined, explained and criticized. After outlining possible improvements and future work that is not yet completed, the paper finishes by presenting the results from fuzzing real-world programs and explains how to recreate them using the provided tool.