Ethchecker: a context-guided fuzzing for smart contracts

Ethereum is the most widely used open-source public chain project, with smart contracts serving as the pattern for developing decentralized applications. The prevalence of attacks against smart contracts has increased in recent years due to the attached amounts of high-value cryptocurrency. Various attacks against smart contracts have caused significant financial losses, amounting to hundreds of millions of dollars. As manual auditing of smart contracts is time-consuming and costly, automatic detection of vulnerabilities is crucial. Existing work does not dig deeper into contextual information contained in the program, which suffers from the difficulty of covering paths with more complex conditions. In this paper, we propose Ethchecker, a smart contract vulnerability detection tool which combines fuzzing and symbolic execution techniques together. Particularly, we propose an analysis module to extract static information from smart contracts. Besides, the tool introduces a genetic algorithm to enlarge code coverage, while considering the contextual information of the code. The results of the experiment show that in terms of F1-score for vulnerability detection, Ethchecker outperforms sFuzz by an average of 21.89% and outperforms Mythril by an average of 12.5%. Furthermore, in the comparison experiments on a dataset consisting of 1000 long smart contract codes (comprising over 3000 instructions), the proposed algorithm can improve the code coverage by 18.56% compared to the random fuzzing algorithm. In addition, we also used Ethchecker to test against 8922 randomly crawled real-world smart contracts. The result demonstrates the stability of this tool.