Neural Network-based Graph Embedding for Cross-Platform Binary Code Similarity Detection

Unknown author(s)

2017 · 499 citations · 16 references

DOIFull text

Open access

Concepts

TL;DR

Cross‑platform binary code similarity detection is crucial for security tasks such as plagiarism, malware, and vulnerability analysis, yet existing graph‑matching methods are slow, inaccurate, and hard to adapt. This work introduces a neural‑network approach that embeds each binary function’s control‑flow graph into a numeric vector, enabling efficient similarity assessment via embedding distance. The authors implement a prototype called Gemini that learns these embeddings and compares them to determine similarity. Gemini achieves substantially higher detection accuracy, reduces embedding generation time by 3–4 orders of magnitude, cuts training time from over a week to 30 minutes–10 hours, and identifies more vulnerable firmware than prior methods.

Abstract

The problem of cross-platform binary code similarity detection aims at detecting whether two binary functions coming from different platforms are similar or not. It has many security applications, including plagiarism detection, malware detection, vulnerability search, etc. Existing approaches rely on approximate graph matching algorithms, which are inevitably slow and sometimes inaccurate, and hard to adapt to a new task. To address these issues, in this work, we propose a novel neural network-based approach to compute the embedding, i.e., a numeric vector, based on the control flow graph of each binary function, then the similarity detection can be done efficiently by measuring the distance between the embeddings for two functions. We implement a prototype called Gemini. Our extensive evaluation shows that Gemini outperforms the state-of-the-art approaches by large margins with respect to similarity detection accuracy. Further, Gemini can speed up prior art's embedding generation time by 3 to 4 orders of magnitude and reduce the required training time from more than 1 week down to 30 minutes to 10 hours. Our real world case studies demonstrate that Gemini can identify significantly more vulnerable firmware images than the state-of-the-art, i.e., Genius. Our research showcases a successful application of deep learning on computer security problems.

References

16