SimCGE: Simple Contrastive Learning of Graph Embeddings for Cross-Version Binary Code Similarity Detection

Fengliang Xia,Guixing Wu, Guochao Zhao, Xiangyu Li

International Conference on Information and Communication Security (ICICS)(2022)

Cited 0|Views22
No score
Binary code similarity detection (BCSD) has many applications in computer security, whose task is to detect the similarity of two binary functions without having access to the source code. Recently deep learning methods have shown better efficiency, accuracy, and potential in BCSD. Most of them reduce losses by the Siamese network, and they ignore some shortcomings of the Siamese network. In this paper, we introduce the idea of contrastive learning into graph neural networks and experimentally demonstrate that the way of training graph models by contrastive learning is significantly better than Siamese. In addition, we found that Principal Neighbourhood Aggregation for Graph Nets (PNA) has the best ability to extract structural information of control flow graph (CFG) among various graph neural networks.
Translated text
Key words
Binary code similarity detection,contrastive learning,graph neural network
AI Read Science
Must-Reading Tree
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined