WCSE 2020 Summer
ISBN: 978-981-14-4787-7 DOI: 10.18178/wcse.2020.06.008

Augmented Fuzzing with Promotion on Numerical Dependence

Sisi Li, Jiaxi Ye, Bin Zhang, Chaojing Tang

Abstract— Coverage-guided fuzzing is an effective technique to find software bugs, it is skilled at eliminating the execution on the repeated paths, thus the fuzzing performance is improved. However, its code coverage-centric design is still insufficient in bug detection. On the one hand, the coverage-based schedule tries its best to reduce the execution on the explored paths even if there are some undetected bugs inside. On the other hand, it mutates the input randomly, which makes it inefficient to generate target testcases. As a result, it may miss some subtle bugs despite it has already explored the vulnerable codes. In this paper, we present a prototype, named ADA, to augment bug detection on the executed paths via a promotion on the numerical dependence of triggering the bug. The numerical dependence means some special numerical conditions that need satisfying to trigger the bug. In a nutshell, our tool allows an intended number of executions on the explored paths, and utilizes a novel approximation search algorithm, to produce the testcases that can satisfy the numerical dependence of triggering the bug and crash the program. Moreover, we leverage a critical field identification method to adjust the mutation, which can help the engine to quickly produce the testcases satisfying the numerical dependence. We implement ADA based on AFL and radare2, and evaluate its performance on some benchmarks. The results demonstrate that ADA outperforms some other state-of-the-art coverage-guided fuzzers in discovering the bugs on the executed paths. It is the proof that our approach is effective. Besides, the experimental results also indicate that the introduced overhead by our approach is within an acceptable scope.

Index Terms— software security, fuzzing, bug detection, numerical dependence, promotion

Sisi Li, Jiaxi Ye, Bin Zhang, Chaojing Tang
College of Electronic Science, National University of Defense Technology , CHINA


Cite: Sisi Li, Jiaxi Ye, Bin Zhang, Chaojing Tang, " Augmented Fuzzing with Promotion on Numerical Dependence " Proceedings of 2020 the 10th International Workshop on Computer Science and Engineering (WCSE 2020), pp. 42-51, Shanghai, China, 19-21 June, 2020.