As blockchain technology continues to advance, smart contracts, a core component, have increasingly garnered widespread attention. Nevertheless, security concerns associated with smart contracts have become more prominent. Although machine-learning techniques have demonstrated potential in the field of smart-contract security detection, there is still a lack of comprehensive review studies. To address this research gap, this paper innovatively presents a comprehensive investigation of smart-contract vulnerability detection based on machine learning. First, we elucidate common types of smart-contract vulnerabilities and the background of formalized vulnerability detection tools. Subsequently, we conduct an in-depth study and analysis of machine-learning techniques. Next, we collect, screen, and comparatively analyze existing machine-learning-based smart-contract vulnerability detection tools. Finally, we summarize the findings and offer feasible insights into this domain.

With the rapid progress of technology, Web 3.0 has emerged as a transformative force in the digital realm. It is characterized by decentralization, user-centric data ownership, and the implementation of cryptographic techniques. Smart contracts, as a core component of Web 3.0, play a pivotal role in driving its evolution by enabling novel functionalities and various application. However, given the substantial financial significance of smart contracts and their inherent transparency, the accessibility of their source code to all opens potential avenues for attackers to identify and exploit vulnerabilities. Therefore, the detection of security vulnerabilities in smart contracts has become significantly important. Existing smart contract vulnerability detection tools mostly rely on expert-defined rules, leading to high false positive rates. To address this problem, this paper proposes an efficient and automated framework that combines Graph and Attention for detecting smart contract vulnerabilities. This framework takes into account the code structure of smart contracts, extracts nodes, and constructs a contract graph, utilizing dataflow to represent the different semantics of variable nodes at different locations. Additionally, a bidirectional multilayer Transformer framework is constructed and trained with our dataset, utilizing the information from the nodes. The framework achieves state-of-the-art levels of $Accuracy$ 92.72%, $Recall$ 82.81%, and $F1_{score}$ 87.54%, respectively. These results show that our framework can effectively detect security vulnerabilities in smart contracts and has the potential to improve their security.