The new methods of machine learning have enabled the identification of criminal activity in the Bitcoin blockchain, such as money laundering and transferring funds to suspicious wallets. Researchers from Elliptic, in collaboration with MIT-IBM Watson Ai Lab, reported on this breakthrough.
During the study, a 26-gigabyte dataset was analyzed, consisting of 122 thousand marked subgraphs within the blockchain, encompassing 49 million units and 196 million transactions. The dataset, labeled by Elliptic2 researchers, provided valuable insights on the connections between wallets and transactions associated with illegal activity on the blockchain.
originally published in July 2019. The project’s objective is to combat financial crimes using machine learning technologies, specifically graph convolutional neural networks (GCN).
Tom Robinson, the chief researcher and co-founder of Elliptic, highlighted that applying machine learning in this context allows for the prediction of whether certain crypto transactions are linked to criminal activities, as opposed to traditional methods that focus on tracking obviously illegal crypto transactions.
The study utilized three classification methods for analyzing the blockchain data: GNN-EG, Sub2VEC, and Glass, which successfully pinpointed numerous crypto accounts potentially involved in illegitimate activities.
Various patterns of cryptocurrency laundering, including the technique known as “Peeling Chain,” were also uncovered. In the Peeling Chain process, an initial amount of cryptocurrency is sent to one address, then portions of it are successively transferred to multiple addresses controlled by the sender, creating a layered effect that obscures the original source of funds.
This method serves to obfuscate the origins of transactions, complicating efforts to trace the flow of funds. While in legitimate scenarios, it can safeguard the privacy of financial transfers, in illegitimate cases, it is used to conceal unlawfully acquired funds.