Graph Attention Network with LSTM for Smart Contract Vulnerability Detection

Abstract

With the advancement of blockchain technology, smart contracts have become a core component of decentralized applications (DApps). However, due to the immutable nature of smart contracts once deployed, it is crucial to detect vulnerabilities before deployment. The increasing complexity of services and the proliferation of service providers pose challenges to manual vulnerability detection. This paper proposes a model that converts the opcode of smart contracts into Control Flow Graphs (CFG) and utilizes a Graph Recurrent Neural Network (GRNN), which is specialized in handling long-term dependencies, to detect vulnerabilities in smart contracts.