HLGAN: Blockchain Account Detection Using Hierarchical Local-Global Attention Network

Blockchain technology plays an important role in the development of the digital economy. As its application fields continue to expand, the transaction security challenges it faces are becoming increasingly prominent and may cause significant economic losses. There are currently a variety of GNN methods used to detect node identities in Ethereum. However, current methods struggle to represent long-range dependencies, thereby posing additional challenges in identifying anonymous accounts on account interaction graphs with heterophily. We propose a novel framework - HLGAN (Hierarchical Local-Global Attention Network), which integrates the local self-attention mechanism and global self-attention mechanism. HLGAN adeptly captures account behavior patterns and effectively integrates global information, thus significantly improving the accuracy of account identification. Experimental results demonstrate that HLGAN exhibits state-of-the-art performance on the Ethereum transaction graph dataset, with significantly better F1-score metrics compared to existing approaches.