Enhancing concurrency vulnerability detection through AST-based static fuzz mutation

Abstract

As multi-threaded and highly concurrent programs are increasingly used, their inherent uncertainty significantly impacts program stability. Traditional testing methods often struggle to effectively detect specific concurrency vulnerabilities because these vulnerabilities are triggered only under particular circumstances, making detection at the vulnerability-triggering level challenging. In view of this, we propose a static fuzz mutation testing method based on Abstract Syntax Tree (AST). This method leverages the fine-grained granularity of ASTs to optimize test suites for concurrency vulnerabilities detection. Initially, We analyze and classify the concurrency vulnerabilities found in Go source code, and generate vulnerability feature mutation operators (mutation operators with concurrency vulnerability feature). Next, we propose static fuzz mutation method and heuristic algorithms at the AST level and apply them to mutators. Ultimately, we screened over 200 code slices from 8 open-source projects for static fuzz mutation testing. The results indicate that introducing vulnerability feature mutation operators improved the number of mutant by approximately 22.15% across various types of concurrency vulnerability samples. This enhancement elevated the probability of triggering concurrency vulnerabilities in the program. After incorporating data from the Go language standard library, experiments further confirmed that our proposed static fuzz mutation testing method can effectively improve the accuracy of concurrency vulnerability detection.