Smart Contract Vulnerability Detection Based on Hybrid Attention Mechanism Model

A smart contract, as an important part of blockchain technology, has attracted considerable interest from both industry and academia. It provides the basis for realization a variety practical applications plays crucial role in ecosystem. While it also holds large number digital assets, frequent occurrence contract vulnerabilities have caused huge economic losses destroyed blockchain-based credit system. Currently, security reliability contracts become new focus research, there are vulnerability detection methods, such traditional tools based on static or dynamic analysis. However, most them often rely expert rules, therefore poor scalability high false negative positive rates. Recent deep learning methods alleviate this issue, but without considering semantic information context source code. To end, we propose hybrid attention mechanism (HAM) model to detect contracts. We extract code fragments code, which key points vulnerability. conduct extensive experiments two public datasets (a total 24,957 contracts). Empirical results show remarkable accuracy improvement over state-of-the art five kinds vulnerabilities, where could achieve 93.36%, 80.85%, 82.56%, 85.62%, 82.19% reentrancy, arithmetic vulnerability, unchecked return value, timestamp dependency, tx.origin, respectively.