Similarity of binaries through re-optimization

We present a scalable approach for establishing similarity between stripped binaries (with no debug information). The main challenge in binary similarity, is to establish similarity even when the code has been compiled using different compilers, with different optimization levels, or targeting different architectures. Overcoming this challenge, while avoiding false positives, is invaluable to the process of reverse engineering and the process of locating vulnerable code.

We present a technique that is scalable and precise, as it alleviates the need for heavyweight semantic comparison by performing out-of-context re-optimization of procedure fragments. It works by decomposing binary procedures to comparable fragments and transforming them to a canonical, normalized form using the compiler optimizer, which enables finding equivalent fragments through simple syntactic comparison. We use a statistical framework built by analyzing samples collected "in the wild" to generate a global context that quantifies the significance of each pair of fragments, and uses it to lift pairwise fragment equivalence to whole procedure similarity.

We have implemented our technique in a tool called GitZ and performed an extensive evaluation. We show that GitZ is able to perform millions of comparisons efficiently, and find similarity with high accuracy.