Abstract
Ethereum fraud detection is crucial for maintaining the security and trustworthiness of blockchain ecosystems. This paper proposes STGraph-FS, a novel framework that combines structural temporal graph neural networks with feature selection techniques for effective Ethereum fraud detection. Our approach captures both the structural properties of transaction networks and their temporal evolution patterns, while simultaneously identifying the most relevant features for fraud detection. Experimental results demonstrate the effectiveness of our method in detecting fraudulent activities in Ethereum networks.