A recent forensic analysis of a dataset graph containing transactions on the Bitcoin blockchain has uncovered clusters linked to illicit activities and money laundering. The analysis detected criminal proceeds sent to a crypto exchange and identified previously unknown wallets belonging to a Russian darknet market.
The study results were the outcome of collaboration between Elliptic and researchers from the MIT-IBM Watson AI Lab.
The dataset, known as Elliptic2, consists of a large graph containing 122K labeled Bitcoin cluster subgraphs within a background graph with 49M node clusters and 196M edge transactions, according to the co-authors of a paper shared with The Hacker News source.
Elliptic2 is an enhancement of the previously released Elliptic Data Set (or Elliptic1) that aims to combat financial crime by utilizing graph convolutional neural networks (GCNs).
The approach involves leveraging blockchain’s pseudonymity along with knowledge of legitimate (e.g., exchange, wallet provider, miner) and illicit services (e.g., darknet market, malware, terrorist organizations, Ponzi scheme) on the network to identify patterns of illicit activity and money laundering.
Tom Robinson, chief scientist and co-founder of Elliptic, explained that machine learning at the subgraph level proved effective in predicting whether crypto transactions constitute proceeds of crime. This method differs from conventional AML solutions that trace funds from known illicit wallets or match patterns of known money laundering practices.
The study, which tested three different subgraph classification methods on Elliptic2, identified subgraphs representing potentially illegitimate activities by crypto exchange accounts. It also traced the source of funds associated with suspicious subgraphs to various entities, including a cryptocurrency mixer, a Panama-based Ponzi scheme, and an invite-only Russian dark web forum.
A closer examination of the subgraphs predicted by the GLASS model revealed familiar cryptocurrency laundering patterns, such as peeling chains and nested services. Robinson noted that identifying these patterns independently using machine learning is a positive sign for combating money laundering.
The research aims to enhance the accuracy and precision of these techniques and expand the analysis to other blockchains in the future.
