Effective Illicit Account Detection on Large Cryptocurrency MultiGraphs

TL;DR

This work tackles illicit account detection in cryptocurrency networks by modeling transactions as a directed multigraph $G=(V,E,\mathbf{X}_E)$ with edge attributes and presenting DIAM, a discrepancy-aware neural framework. DIAM combines Edge2Seq, which learns node representations from sequences of incoming/outgoing edge attributes via GRUs, with Multigraph Discrepancy (MGD), a directed, attention-guided message-passing module that propagates both neighbor features and their discrepancies. Trained end-to-end with binary cross-entropy, DIAM outperforms 15 baselines across four large Bitcoin/Ethereum datasets, achieving state-of-the-art F1 and AUC scores (e.g., $F1=96.55\%$ on Bitcoin-L, outperforming the runner-up at $83.92\%$). The approach effectively leverages edge sequences and cross-edge discrepancies, offering a scalable solution for illicit-account detection with potential applicability to other transaction networks beyond cryptocurrency.

Abstract

Cryptocurrencies are rapidly expanding and becoming vital in digital financial markets. However, the rise in cryptocurrency-related illicit activities has led to significant losses for users. To protect the security of these platforms, it is critical to identify illicit accounts effectively. Current detection methods mainly depend on feature engineering or are inadequate to leverage the complex information within cryptocurrency transaction networks, resulting in suboptimal performance. In this paper, we present DIAM, an effective method for detecting illicit accounts in cryptocurrency transaction networks modeled by directed multi-graphs with attributed edges. DIAM first features an Edge2Seq module that captures intrinsic transaction patterns from parallel edges by considering edge attributes and their directed sequences, to generate effective node representations. Then in DIAM, we design a multigraph Discrepancy (MGD) module with a tailored message passing mechanism to capture the discrepant features between normal and illicit nodes over the multigraph topology, assisted by an attention mechanism. DIAM integrates these techniques for end-to-end training to detect illicit accounts from legitimate ones. Extensive experiments, comparing against 15 existing solutions on 4 large cryptocurrency datasets of Bitcoin and Ethereum, demonstrate that DIAM consistently outperforms others in accurately identifying illicit accounts. For example, on a Bitcoin dataset with 20 million nodes and 203 million edges, DIAM attains an F1 score of 96.55%, markedly surpassing the runner-up's score of 83.92%. The code is available at https://github.com/TommyDzh/DIAM.

Figures (6)

• Figure 1: A directed multigraph with edge attributes.
• Figure 2: The DIAM framework with an input transaction network modeled as a directed multigraph with edge attributes.
• Figure 3: Compare the MGD module with other GNN layers.
• Figure 4: Performance Comparison with Varying $T_{max}$
• Figure 5: Varying illicit ratio (%) on all datasets.