Smart Contract Vulnerability Detection Based on Machine Learning

Abstract

Smart contracts are among the most important applications of blockchain technology and are vulnerable to network attacks, leading to significant financial losses. Thus, smart contract vulnerability detection has become an important research field. Currently, there are problems with limited detection types and low detection efficiency in smart contract vulnerabilities. It is particularly crucial to achieve high detection efficiency and cover a wide range of vulnerability types in smart contract vulnerabilities. This paper builds a dataset named SC-4, which consists of over 3000 data of four common vulnerabilities in smart contracts: Reentrancy, Transaction Order Dependence, Unchecked-Send, and Unhandled-Exceptions. We propose a smart contract vulnerability detection model based on an improved gated recurrent unit (GRU) and random forest (RF) fusion algorithm. The experimental results show that the proposed algorithm can accurately identify four types of smart contract vulnerabilities, and the accuracy can reach 98.47%. The number of vulnerability types detected has also increased compared to that in previous research.