SymSem: Symbolic Execution with Time Stamps for Deobfuscation

Abstract

Code virtualization technique obfuscates programs by transforming original code to self-defined bytecode in a different instruction architecture. It is widely used in obfuscating malware for its ability to render normal analysis ineffective. Using symbolic execution to assist in deobfuscating such programs turned to be a trend in recent research. However, we found many challenges that may lead to semantic confusion in previous symbolic execution technique, and proposed a novel symbolic execution technique enhanced by time stamps to tackle these issues. For evaluation, we implemented it as a prototype of SymSem and deobfuscated programs protected by popular virtual machines. The results indicate that our method is able to accurately recover the semantics of obfuscated function trace.