Predicting Risk Links in Ethereum Networks Using Transaction Behavior Synchronization

With its rapid development and distinctive architecture, blockchain has become a fundamental infrastructure for heterogeneous data interaction and trusted sharing. It plays an important role in applications such as Cyber-Physical-Social Systems (CPSS). However, phishing scams, a typical form of cybercrime, have recently spread to the blockchain ecosystem, causing substantial financial losses in Ethereum. Therefore, there is an urgent need for methods to predict and identify phishing risk links to ensure the long-term security and stability of blockchain-driven CPSS and financial systems. Most existing studies focus on static phishing address identification, and the early warning capability for potential risky transactions remains limited. In addition, many existing methods suffer from weak interpretability and high computational cost. In this paper, we propose a method, Behavioral Synchronization of Transactions Enables Risk Prediction (BST-RP), which models behavioral sequences over time intervals and quantifies their coupling by incorporating both transaction counts and values. The BST-RP captures the synchronization and similarity of user interactions and can identify potential recurrent phishing risk transactions. Experiments on public and crawled real-world datasets show that BST-RP achieves an average AUC of 0.9228 across six transaction graphs, improving over baseline methods by 0.0351–0.7133, demonstrating strong robustness and performance in predicting phishing risk links.