A function-level selective code obfuscation method based on LLVM IR

Abstract

A function-level selective code obfuscation method based on LLVM IR relates to the fields of software engineering and software security. This method relies on the Low-Level Virtual Machine Compiler Infrastructure (LLVM) and its obfuscation extension framework, Obfuscator-LLVM. Addressing the problem of high overhead caused by applying uniform obfuscation to the entire program in existing code obfuscation methods, it proposes a differentiated obfuscation scheme that allocates obfuscation intensity at the function level. The method includes: obtaining the function-level intermediate representation of the program; extracting function-level static structural features; constructing obfuscation level labels and training a function-level obfuscation level prediction model; predicting the obfuscation level of functions in the program to be protected; writing the prediction results into function metadata; and selectively performing corresponding obfuscation transformations based on the metadata during the compilation stage to generate the target obfuscated program. This invention can enhance the protection effect of critical functions while reducing the additional overhead of non-critical functions, thereby achieving a balance between program protection capabilities and engineering costs.

Images