Researchers from Elliptic, IBM Watson, and MIT used AI to detect money laundering on the Bitcoin blockchain.
Back in 2019, blockchain analytics firm Elliptic published an MIT- Published research results with IBM Watson AI Lab.
Now, the partners have published new research applying the new technology to a much larger dataset containing nearly 200 million transactions. Rather than identifying transactions made by fraudsters, the machine learning model was trained to identify “subgraphs,” chains of transactions that represent the Bitcoins being laundered.
Identifying these subgraphs rather than illicit wallets allows researchers to focus on the more general “multi-hop” laundering process rather than the on-chain behavior of specific illegal actors.
The researchers tested the technology in collaboration with a cryptocurrency exchange. Of the 52 money laundering subgraphs that were predicted and ended in exchange deposits, 14 were received by users who had already been flagged as related to money laundering.
On average, fewer than 1 in 10,000 accounts were flagged in this way, “suggesting that the model is performing very well,” the team said. The researchers are now making the underlying data publicly available.
Elliptic said: “This new research shows that AI techniques can be applied to blockchain data to identify previously hidden patterns of illicit wallets and money laundering.
“This is made possible by the inherent transparency of blockchain and shows that crypto assets are not a haven for criminals and are much more suitable for AI-based financial crime detection than traditional financial assets.”
