Multifuzz: a Collaborative Fuzzing Framework Based on Concolic Execution

Résumé fourni par la source

Fuzzing is an efficient vulnerability discovery technique, but it has limitations in exploring the depth of program paths. Concolic Execution performs outstandingly in path coverage, yet its execution speed is relatively slow. Hybrid fuzzing combines the advantages of both, improving the comprehensiveness and efficiency of testing. Collaborative fuzzing further enhances the testing effectiveness by integrating multiple testing engines. This paper studies the impact of concolic execution on the collaborative fuzzing framework and proposes a new framework-MultiFuzz. This framework contains an independent concolic execution module, which improves flexibility. It also incorporates a priority-based seed synchronization strategy and a seed redundancy detection algorithm to further improve testing efficiency. Firstly, the priority-based seed synchronization strategy improves the coverage rate in the early stage of testing. The experiments show that the coverage rate increases by 5.64 % in the first two hours. Secondly, the seed redundancy detection algorithm improves the resource utilization of concolic execution by removing redundant seeds, reducing the redundant seed rate from 33.24 % to 14.19 %. Finally, the experimental results of the MultiFuzz framework indicate that, on UNIFUZZ, compared with EnFuzz and autofz, MultiFuzz has a maximum increase of 21.14 % and 7.46 % in coverage rate, and a maximum increase of 61.17 % and 24.81 % in the number of unique bugs discovered.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.