EffiTaint: Boosting Sensitive Data Tracking with Accurate Taint Behavior Modeling and Efficient Access Path Optimization

Haocheng Li, Haipeng Qu, Gaozhou Wang, Xiangjian Ge

Abstract: The increasing costs of data breaches have made enhancing security and privacy measures more critical than ever. Static taint analysis has become a key technique for tracking sensitive data propagation. However, existing tools such as FlowDroid and Tai-e face significant challenges related to performance, precision, and the handling of complex taint propagation behaviors, such as those involving arrays and taint elimination. In response, we present EffiTaint, a novel static taint analysis tool. EffiTaint constructs the Pointer Flow Graph (PFG) on demand and employs a taint-driven selective analysis strategy, avoiding the overhead of constructing a complete call graph. EffiTaint enhances the modeling of taint propagation within arrays and taint elimination in program control flows, thereby effectively reducing the generation of erroneous sensitive data flows. Finally, EffiTaint optimizes the source and sink matching strategy using a method extension algorithm, significantly improving the ability to detect taint flows. Benchmarking results demonstrate that EffiTaint achieves high precision (97.5%), high recall (96.7%), and low overhead (average runtime of 26.93 seconds, memory usage of 0.67 GB), making it an effective solution for enhancing privacy and security in Java programs.