Memory Error Detection Based on Dynamic Binary Translation

Abstract

Dynamic memory error detection technology is widely utilized in locating and mitigating memory-corruption vulnerabilities. High false negative in existing technologies impairs the effectiveness of detection, especially when the detector has no source code. We proposed a dynamic memory error detection method based on dynamic binary translation. It realized the detection by combining a call stack tracing method and an IR language level memory error detection method. Experiments show that the detection success rate of this method in the heap and stack memory destruction is more than 80%.