Bi-directional Taint Flow Analysis: A High-precision Static Detection Approach for Java Deserialization Vulnerabilities
Haihua Liu, Yanrong Lu · 2025
The Java deserialization vulnerability represents a significant security threat to enterprise applications, enabling attackers to execute malicious code through crafted serialized data. Current detection approaches struggle with accurately identifying complex exploit chains and effectively capturing comprehensive data flow paths through traditional one-way taint analysis, resulting in false positives and missed vulnerabilities. Our proposed bi-directional taint flow analysis method innovatively combines forward taint propagation with reverse sensitivity analysis to precisely identify potential deserialization exploit chains through intersection analysis. Experimental results demonstrate that our SerialVulnScanner implementation achieves a 74.3% detection rate on mainstream Java components—6.9 percentage points higher than existing tools—with particular effectiveness in handling complex code structures and atypical exploit chains, thereby providing enhanced security for Java applications.