The collaboration concluded that transaction analytics could be a key tool in helping banks and PSPs identify financial crime patterns in real-time retail payment systems. Project Hertha tested the application of AI techniques to help spot complex criminal activity. The project, whose findings are contained in a 30-page report, took its name from prolific British inventor and suffragette Hertha Ayrton.
The experiments helped banks and PSPs find 12% more illicit accounts. Payment system analytics proved to be important for spotting novel financial crime patterns, helping achieve a 26% increase in this regard.
The project was put in motion by the need to protect payment systems from financial crime while upholding user privacy, which has been a challenge for the industry. To escape detection, fraudsters operate in complex networks which include many accounts across multiple financial institutions.
Tests were conducted using a simulated synthetic transaction dataset, which included information for 1.8 million bank accounts and 308 million transactions. The dataset was built by an AI model trained to simulate realistic transaction patterns, and it represented an ecosystem of retail payments in a single jurisdiction.
The results are promising, but also showcase limits to the application and system analytics effectiveness. BoE and BIS concluded that the introduction of a similar solution would raise practical, regulatory, and legal issues.
Payment system analytics proved most effective when targeted at identifying more complex schemes involving different accounts from multiple banks and PSPs. For some schemes, the detection precision doubled. For best results, the algorithm needs to be trained on confirmed past cases and supervised.
The collaboration with BoE comes as BIS prepares to make leadership changes. Its Innovation Hub continues to work on fintech and central bank developments amid the transition, focusing on tools to optimise the functioning of the global financial system.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now