Researchers from Elliptic, IBM Watson, and MIT have leveraged AI to uncover money laundering activities on the Bitcoin blockchain.
In 2019, Elliptic, a blockchain analytics firm, collaborated with the MIT-IBM Watson AI Lab to showcase how a machine learning model could identify Bitcoin transactions linked to illicit activities, including those by ransomware groups and darknet marketplaces.
Recently, the collaborators released new research applying advanced techniques to a significantly larger dataset, encompassing nearly 200 million transactions. Instead of focusing solely on transactions related to illicit actors, the new model was designed to detect "subgraphs"—chains of transactions indicative of money laundering.
By identifying these subgraphs, the researchers could analyze the broader "multi-hop" laundering process rather than just the on-chain actions of specific illicit entities. They partnered with a cryptocurrency exchange to test their method, and out of 52 money laundering subgraphs identified that culminated in deposits to the exchange, 14 were linked to users already flagged for money laundering activities.
This indicates that the model operates effectively, with less than one in 10,000 accounts raising flags. The research team is now making their underlying data publicly accessible.
Elliptic stated, "This novel work demonstrates that AI methods can be effectively applied to blockchain data to uncover illicit wallets and money laundering patterns that were previously obscured." They emphasized that the transparency inherent in blockchains enables more efficient AI-based detection of financial crimes compared to traditional financial assets.
Explore the new challenges and opportunities that artificial intelligence brings to the banking sector at Finextra’s inaugural NextGenAI conference on November 26, 2024. Register your interest here.