Blockchain researchers have successfully utilized AI technology to detect potential instances of money laundering linked to Bitcoin transactions. A collaborative effort involving experts from Elliptic, IBM Watson, and MIT resulted in the development of a machine learning model that could identify suspicious activities on the Bitcoin blockchain.
In a study conducted back in 2019, Elliptic and the MIT-IBM Watson AI Lab demonstrated the effectiveness of machine learning in identifying illicit Bitcoin transactions, particularly those associated with ransomware groups or darknet marketplaces. Building upon this research, the team recently examined a much larger dataset comprising nearly 200 million transactions.
Rather than pinpointing individual transactions involving illicit actors, the focus shifted towards identifying subgraphs – chains of transactions indicative of Bitcoin being laundered. This approach allowed researchers to delve deeper into the multi-hop laundering process rather than solely concentrating on specific illicit entities.
Collaborating with a cryptocurrency exchange, the researchers tested their methodology by predicting 52 money laundering subgraphs, with 14 of them concluding in deposits to users previously flagged for money laundering activities. This high success rate – less than one in 10,000 accounts being flagged – underscores the model’s efficacy in identifying potential instances of illicit financial practices.
Elliptic emphasized the significance of this groundbreaking work, highlighting the transparency of blockchains that enables the application of AI techniques for detecting illicit wallets and money laundering patterns. The team believes that cryptoassets, contrary to misconceptions, are conducive to AI-driven financial crime detection, surpassing traditional financial assets in terms of visibility and traceability.
