A dynamic marking method for implicit information flow in dynamic taint analysis

Abstract

Dynamic taint analysis is an important technique for tracking information flow in software and it has been widely applied in the field of software testing, debugging and vulnerability detection. However, most of the dynamic taint analysis tools only handle explicit information flow, while ignoring the implicit information flow, resulting in a large number of false negative errors. Considering this situation, we present a dynamic marking method for implicit information flow, to handle a specific type of control-dependence. The method can identify and propagate implicit data during runtime, thus increasing the coverage of the tested program. we also propose pipeline, integrating our method in the process of dynamic taint analysis. Pipeline is implemented on the base of the dynamic taint analysis framework avalanche, and is designed to detect vulnerabilities in binary programs. In the studies, we applied the tool to 5 applications from some open-source projects, and it has effectively located and propagated the specific kind of implicit information flow.