SigTran: Signature Vectors for Detecting Illicit Activities in Blockchain Transaction Networks
Farimah Poursafaei, Reihaneh Rabbany, Z. Zilic
Pacific-Asia Conference on Knowledge Discovery and Data Mining
Abstract
Cryptocurrency networks have evolved into multi-billion-dollar havens for a variety of disputable financial activities, including phishing, ponzi schemes, money-laundering, and ransomware. In this paper, we propose an efficient graph-based method, SigTran, for detecting illicit nodes on blockchain networks. SigTran first generates a graph based on the transaction records from blockchain. It then represents the nodes based on their structural and transactional characteristics. These node representations accurately differentiate nodes involved in illicit activities. SigTran is generic and can be applied to records extracted from different networks. SigTran achieves an F_1 score of 0.92 on Bitcoin and 0.94 on Ethereum, which outperforms the state-of-the-art performance on these benchmarks obtained by much more complex, platform-dependent models.