Investigation of advancing LLMs model for smart contract vulnerabilities detection

Abstract

In the field of cyber security, while blockchain technology is renowned for its robust security, the blockchain smart contracts suffer of various vulnerabilities that attackers can easily exploit to launch attacks, resulting in irreversible losses. Therefore, ensuring the security of smart contracts is crucial given the widespread adoption of this technology in our society. At present, although there are many traditional methods used for vulnerability detection, these methods generally have certain limitations. With the rapid development of deep learning, AI (particularly generative AI) has gradually become mainstream in the field of software engineering vulnerability detection, offering new opportunities for enhancing smart contract security. In this paper, we investigate how advanced large language models (LLMs) can be used to tackle vulnerability detection, not only to identify vulnerability but also to provide remediation suggestions for fixing vulnerable smart contracts. We introduce, through a new guideline-based Framework, a suitable application process for using LLMs in smart contract vulnerability detection activity using less computing resource.