SeedUI: Understanding Initial Seeds in Fuzzing

Greybox fuzzing, a widely used dynamic testing technique, iteratively tests the software using semi-randomly generated inputs that aim at reaching more code at runtime. It relies on a set of initial inputs (initial seeds) to bootstrap the fuzzing process. These initial seeds highly influence the fuzzing performance in terms of runtime coverage or finding bugs. This paper introduces SeedUI, a tool that helps in visualizing and understanding the performance of initial seeds across multiple fuzzing campaigns. It provides five different views that help a user for an in-depth initial seed analysis. In our previous work, we extracted a visualization task taxonomy for greybox fuzzing by interviewing 33 fuzzing experts and analyzing the responses. We evaluate SeedUI along with four existing tools: VisFuzz, FuzzSplore, FuzzInspector, and Ijon UI, in terms of their support to the taxonomy. Our findings indicate that SeedUI complements existing tools by addressing 9 tasks in the taxonomy that are not supported by them.Demonstration video: https://youtu.be/qpPjutmIcTs Artifacts: https://github.com/secure-software-engineering/SeedUI