On the (in)efficiency of fuzzing network protocols

Abstract

AbstractFuzzing is a widely used and effective technique to test software. Unfortunately, certain systems, including network protocols, are more challenging to fuzz than others. An important complication with fuzzing network protocols is that this tends to be a slow process, which is problematic as fuzzing involves many test inputs. This article analyzes the root causes behind the inefficiency of fuzzing network protocols and strategies to avoid them. It extends our earlier work on network protocol fuzzers, which explored some of these strategies, to give a more comprehensive overview of overheads in fuzzing and ways to reduce them.