Prompt Fuzzing for Fuzz Driver Generation

arXiv:2312.17677v2 Announce Type: replace
Abstract: Crafting high-quality fuzz drivers not only is time-consuming but also requires a deep understanding of the library. However, the state-of-the-art automatic fuzz driver generation techniques fall short of expectations. While fuzz drivers derived from consumer code can reach deep states, they have limited coverage. Conversely, interpretative fuzzing can explore most API calls but requires numerous attempts within a large search space. We propose PromptFuzz, a coverage-guided fuzzer for prompt fuzzing that iteratively generates fuzz drivers …

art arxiv automatic can code consumer consuming cs.cr cs.se driver drivers fuzz fuzzing high library prompt quality state states techniques understanding