Efficient Vulnerability Detection in Solidity Ethereum Smart Contracts Through Bytecode Analysis Using Customized CNN

Abstract

Smart contracts are an integral part of blockchain networks, enabling automation and security of transactions. But undetected smart contract vulnerabilities can be exploited severely. Manual code review and static analysis, which are traditional methods for detecting vulnerabilities, tend to be time-consuming and do not possess the capability to identify sophisticated flaws. This work proposes a new method using a tailored Convolutional Neural Network (CNN) to examine Ethereum Solidity smart contract bytecode to provide effective and strong vulnerability detection. Our model is highly accurate and efficient, with 87.75 % accuracy, 86.38 % precision, 87.09 % recall, and an F1-score of 86.22 % on the Slither audited smart contract dataset. This innovation offers an efficient and scalable solution for Ethereum-based smart contracts security.