HGAT: smart contract vulnerability detection method based on hierarchical graph attention network

Abstract With the widespread use of blockchain, more and smart contracts are being deployed, their internal logic is getting sophisticated. Due to large false positive rate low detection accuracy most current methods, which heavily rely on already established criteria, certain additionally call for human secondary detection, resulting in efficiency. In this study, we propose HGAT, a hierarchical graph attention network-based model, order address aforementioned issues as well shortcomings contract vulnerability approaches. First, using Syntax Tree (AST) Control Flow Graph, functions abstracted into code graphs (CFG). Then abstract each node subgraph, extract features, utilize mechanism GAT, splice obtained vectors form features line statements these detect contracts. To create test data assess leverage open-source sample dataset. The findings experiment indicate that method can identify vulnerabilities quickly precisely than other techniques.