Topological Anomaly Detection in Dynamic Multilayer Blockchain Networks

Motivated by the recent surge of criminal activities with cross-cryptocurrency trades, we introduce a new topological perspective to structural anomaly detection in dynamic multilayer networks. We postulate that anomalies underlying blockchain transaction graph are composed multiple layers likely also be manifested anomalous patterns network shape properties. As such, invoke machinery clique persistent homology on graphs systematically and efficiently track evolution and, as result, detect changes topology geometry. develop persistence summary for networks, called stacked diagram, prove its stability under input data perturbations. validate our framework application networks from Ethereum Blockchain Ripple Credit Network, demonstrate PD approach substantially outperforms state-of-art techniques.