GCVAT: A Hybrid Graph Convolution and Attention Model for Smart Contract Vulnerability Detection

Abstract

Smart contracts have become an integral component of blockchain technology, enabling automated and decentralized transactions. However, their increasing adoption has exposed critical security vulnerabilities, with the reentrancy vulnerability being one of the most prominent threats. This vulnerability arises when external calls to other contracts are made before the completion of a transaction, allowing malicious actors to exploit the contract’s state. In this paper, we propose an efficient and novel solution that will be able to detect both reentrancy and infinite loop vulnerabilities, called GCVAT. Our model analyzes the interactions and dependencies among various components of smart contracts through a series of mechanisms. The GCVAT model is a revolutionary hybrid model that is created by combining the capabilities of a GCN and a GAT model. We present a comprehensive evaluation of our GCVAT model through simulations, showcasing its effectiveness in identifying vulnerabilities in a range of smart contracts. The results demonstrate significant improvements in the detection of accuracy and scalability, making our approach a valuable contribution to ongoing efforts to secure smart contracts. Ultimately, this research aims to foster greater trust in blockchain applications by mitigating the risks associated with reentrancy and infinite loop vulnerabilities.