A Method of Handling Control Dependence in Dynamic Taint Analysis

Dynamic taint analysis technique plays an important role in tracking the input data´s trace in an executing program, and it has been widely used in program analysis method, such as information-flow analysis and dynamic slicing. However, most of the tools implementing dynamic taint analysis only consider the data dependence propagation and neglect the control dependence. In this paper, we presents a dynamic method of handling specific control dependence problems for binary programs. We use the framework Valgrind to perform instrumentation into binary programs. By analyzing the characteristic of the valgrind intermediate representation and the execution binary program traces, we propose the method to make propagation policies targeted on the control-dependence problems. In the studies, we apply our method into the process of dynamic taint analysis and it shows that our method has effectively mark the variables influenced by control dependence.