As a test of the resulting AI tool, the researchers verified their results with a cryptocurrency exchange (which the paper does not name) by identifying 52 suspicious chains of transactions that had ultimately flowed to that exchange. It turned out that the exchange had already flagged 14 of the accounts that had received those funds for suspected illicit activity, including 8 that it had flagged as being associated with money laundering or fraud, based in part on the “know your customer” information it had requested. of the account owners. Despite not having access to know-your-customer data or any information about the origin of the funds, the researchers’ AI model matched the findings of the exchange’s own researchers.
Correctly identifying 14 out of 52 of those customer accounts as suspicious may not seem like a high success rate, but researchers note that only 0.1 percent of the exchange’s accounts are flagged for possible money laundering overall. They argue that their automated tool had essentially reduced the search for suspicious accounts to more than one in four. “Going from ‘one in a thousand things we look at will be illicit’ to 14 out of 52 is a crazy change,” says Mark Weber, one of the paper’s co-authors and a member of MIT’s Media Lab. “And now investigators are going to investigate the rest to see, wait, did we miss something?”
Elliptic says it has already been privately using the AI model in its own work. As further evidence that the AI model is producing useful results, the researchers write that analyzing the source of funds for some suspicious transaction chains identified by the model helped them uncover Bitcoin addresses controlled by a Russian dark web marketplace, a cryptocurrency “mixer” designed to obfuscate the trail of bitcoins on the blockchain and a Panama-based Ponzi scheme. (Elliptic declined to identify any of those alleged criminals or services by name and told WIRED that it does not identify the targets of the ongoing investigations.)
Perhaps more important than the practical use of the researchers’ own AI model, however, is the potential of Elliptic’s training data, which the researchers have published on Google-owned data science and machine learning community site Kaggle. “Elliptic could have kept this,” says MIT’s Weber. “Instead, here was an open source spirit of contributing something to the community that will allow everyone, even your competitors, to be better at fighting money laundering.” Elliptic notes that the data it published is anonymous and does not contain any identifiers of the owners of the Bitcoin addresses or even the addresses themselves, only the structural data of the “subgraphs” of the transactions it labeled with its suspected laundering qualifications. of money.
That huge trove of data will certainly inspire and enable much more AI-focused research into Bitcoin money laundering, says Stefan Savage, a computer science professor at the University of California, San Diego, who was an advisor to the Lead author of a fundamental study on Bitcoin tracking. article published in 2013. However, he maintains that the current tool does not seem likely to revolutionize cryptocurrency anti-money laundering efforts in its current form, but instead serves as a proof of concept. “I think an analyst is going to have difficulties with a tool that is gentle sometimes it’s the right thing to do,” Savage says. “I see this as a trailer that says, ‘Hey, there’s something here. More people should work on this.’”
