Cryptocurrency exchange Coinbase has secured a patent for a self-learning system that identifies and flags non-compliant user accounts, according to a filing with the U.S. Patent and Trademark Office published on 19 November.
The document describes an automated system with a scoring model, which would aim to detect non-compliant accounts, especially those suspected of illegal activities.
It will use a variety of factors to assign an overall compliance score to each account, and then compare the scores to find those that fail the compliance standards.
While some of the factors are calculated by the exchange itself, others are user-inputted or executed data points, and will include user age, account balance, transaction volume, location, verification history and the number of devices with access.
The system will also measure “the level of due diligence” performed on accounts, and the history of compliance reviews, in order to determine whether an account is “bad or good”.
The filing reads:
“An investigator may be able to determine whether an account is being used for illicit activities by doing research on the parties of the transaction who receive or send payment and determining whether such parties are regularly involved in illicit activities. It may for example be relatively easy to determine that a party sending or receiving payment is in the business of conducting online services that may be illegal.”
The system was also designed to learn and constantly update its compliance model from the data that it collects, and the accounts it is flagging.
Though accounts that are flagged as “bad” are suspended and reported to law enforcement authorities if there is a transaction for more than $2,000, investigators will have an override function that will enable them to bypass the suspension.
Yesterday, blockchain security company CipherTrace released a report, titled “Cryptocurrency Anti-Money Laundering Report, 2019 Q3”, which claimed that the total volume of cryptocurrency-related fraud and theft resulted in losses worth $4.4 billion in 2019.
The firm also analyzed the 120 most popular cryptocurrency exchanges’ Know Your Customer (KYC) and Anti-Money Laundering (AML) requirements, and concluded that two-thirds of exchanges lack strong KYC policies.