Money-mule networks move criminal funds through many small transfers, each of which appears normal on its own. A new paper on the arXiv argues that what marks an account as a mule lies outside any single transfer — it is found in the counterparties and addresses connected to that account.
The study frames money-mule detection as a relational learning problem and uses the RBIH NFPC dataset to explore this approach. Rather than examining transactions in isolation, the authors emphasize the importance of the account's network of relationships.
This suggests that conventional per-transaction screening may be insufficient. Detecting mules likely requires models that incorporate the structure of the surrounding network, not just the features of individual payments.