Smart Contract Vulnerability Detection Based on Critical Combination Path and Deep Learning

Abstract

Ethereum is currently one of the most popular blockchain platforms. Smart contracts are an important part of blockchain. Because developers lack understanding of contract security and the huge value of contracts themselves, contracts are often attacked. Therefore, how to effectively detect smart contract vulnerabilities has become a crucial issue. This paper uses deep learning to detect vulnerabilities, which can get rid of dependence on expert experience. In order to solve the problem of poor detection effect caused by excessive noise, this paper proposes a vulnerability detection technology based on critical combination path and deep learning. The critical combination path only contains code related to vulnerabilities, eliminating many invalid codes, thus greatly reducing the impact of noise. At the same time, by analyzing the characteristics of assembly code, a normalization method is proposed to remove many homogeneous codes. The normalized critical combination paths are then vectorized using SimHash, and then converted to grayscale images for classification using a neural network. The experimental results show that the proposed scheme is effective.