Malware programs that incorporate trigger-based behavior initiate malicious activities based on conditions satisfied only by specific inputs. State-of-the-art malware analyzers discover code guarded by triggers via multiple path exploration, symbolic execution, or forced conditional execution, all without knowing the trigger inputs. We present a malware obfuscation technique that automatically conceals specific trigger-based behavior from these malware analyzers. Our technique automatically transforms a program by encrypting code that is conditionally dependent on an input value with a key derived from the input and then removing the key from the program. We have implemented a compiler-level tool that takes a malware source program and automatically generates an obfuscated binary. Experiments on various existing malware samples show that our tool can hide a significant portion of trigger based code. We provide insight into the strengths, weaknesses, and possible ways to strengthen current analysis approaches in order to defeat this malware obfuscation technique.
"Impeding Malware Analysis Using Conditional Code Obfuscation"
Monirul Sharif, Andrea Lanzi, Jonathon Giffin and Wenke Lee
In the Proceedings of the 15th Annual Network and Distributed System
Security Symposium (NDSS'08), San Diego, CA, February 2008