Bitcoin Money Laundering Now Detectable Using AI, UK Gov Says Crypto Firms Pose Greatest Risk
Detecting crypto money laundering can be a challenge. Photo by Oli Woodman on Unsplash.
Key Takeaways
- Researchers have proposed a new method of identifying crypto money laundering with AI.
- Unlike approaches that rely on known money laundering activity, the proposed technique flags suspicious patterns by analyzing clusters of blockchain addresses.
- AI tools could become increasingly important for crypto firms that are under pressure to beef up their AML measures.
According to the UK Treasury’s annual anti-money laundering report, the crypto sector poses one of the greatest risks of being exploited by money launderers. So how can businesses that deal with cryptocurrency avoid being implicated?
In a paper published on Wednesday, May 1, researchers from MIT, IBM and Elliptic proposed a new technique that uses graph neural networks (GNNs) to identify the “shape” of illicit Bitcoin transactions. For crypto firms and law enforcement agencies, the technology could prove a valuable weapon in the war on money laundering.
Artificial Intelligence in AML
Modern financial service providers incorporate a range of AI tools in their anti-money laundering (AML) arsenal.
For banks, AI is used for transaction monitoring to identify patterns of suspicious activity. But as more financial institutions embrace crypto, they are turning their attention to on-chain transactions too.
In both realms, the underlying concept is the same. Using past transaction data, with activity that is known to be implicated in money laundering labeled as such, AI models can learn to spot patterns of suspicious behavior.
This model has proven very successful for banks, which have access to suspicious activity reports going back decades with which they can train AI.
But cryptocurrencies are much newer. Regulations that require platforms to collect and share information on suspected money laundering incidents are an even more recent development. As such, there is much less of the annotated data needed to train AI to scan for suspicious transactions.
Overcoming the Lack of Labeled Transaction Data
With less information about known money laundering paths, researchers have turned their attention to what they do know something about: addresses associated with illicit activity.
The new GNN-based technique works by defining clusters of addresses associated with criminal actors such as darknet markets or fraud schemes. The next step is to identify clusters linked to crypto exchanges that money launderers use to cash out their illicit proceeds.
Mapping Money Laundering Networks
According to the research paper “a path on the blockchain connecting an illicit cluster to a licit cluster without a change of ownership of the funds likely represents the activity of money laundering by a criminal person or organization.”
By modeling the connections between illicit and licit clusters as subgraphs of the larger blockchain network, researchers say they have identified a distinct subgraph “shape” that machine learning models can identify.
Of the subgraphs deemed suspicious using this technique, at least sixty received funds from crypto mixers; at least twenty received funds from a node believed to be a Russian darknet market; a further 2 received funds from a Panama-based ponzi scheme.
Crypto Firms Implicated
The latest research into AI-based techniques for on-chain transaction monitoring comes as crypto firms are under growing pressure to beef up their AML measures.
In the UK, the Treasury report notes that businesses dealing with crypto assets “remained particularly vulnerable to financial crime,” despite the FCA’s crackdown on non-compliant platforms.
The report also flags “retail banking (including payments), wholesale banking [and] wealth management,” and doesn’t indicate that the crypto sector is any more implicated in money laundering.
Nevertheless, the UK’s financial regulator has diverted significant resources to supervising crypto firms. Although just a quarter of the money laundering cases it opened last year involve crypto asset businesses, 30% of the Financial Conduct Authority’s financial crime experts are employed overseeing the sector. Facing such scrutiny, better AML tools can’t come fast enough.
