Exploring Compiler Optimization Space for Control Flow Obfuscation
Hameeza Ahmed,Muhammad Faraz Hyder,Muhammad Fahim ul Haque,Paulo Cesar Santos
DOI: https://doi.org/10.1016/j.cose.2024.103704
IF: 5.105
2024-01-11
Computers & Security
Abstract:Code obfuscation is a promising technique for securing software and protecting it from adversaries. The objective is to harden the exploitation of security vulnerabilities for the attacker as well as launching of successful attacks. Obfuscation can be classified into layout, data, and control flow obfuscation. Control flow obfuscation impedes the understanding of the application logic by making it complicated to determine the actual control flows. Although numerous control flow methods exist in the literature, the role of existing compiler optimizations has just been discovered. This paper is the first one that explores the existing optimization space of LLVM compiler for obfuscating code. Our techniques optimally explore the native compiler's optimizations to improve the original code performance and reduce memory space with no disruptive efforts, tools, or extra costs. In the CBench benchmark suite, our work is able to improve 246%, 143%, and 468% in cyclomatic complexity, program length, and implementation effort, respectively, compared to unobfuscated code. Therefore, instead of inventing new obfuscation tools, the existing compiler optimizations can easily be used to obfuscate control flows, saving the overall cost and efforts.
computer science, information systems