Laundered money isn’t stashed away quietly, but has to keep on the move. | Photo: Keystone/Gabriele Putzu

Criminals have all kinds of tricks to launder their dishonestly acquired money. They might distribute it into different accounts, then merge it again. Or they might move it around in a loop through several different banks. Manoeuvres like this aren’t obvious when you’re looking at tabular account statements.

The patterns that are typical of such transactions can nevertheless be detected by converting them into a graph that depicts money flows like a public transport map. A team from IBM Research Europe Switzerland, based in Zurich, has now developed special algorithms for this kind of analysis, and found when testing them that they were able to recognise suspicious money movements significantly better than with other methods. What’s also new is the fact that this analysis can be carried out in real time.

According to the study’s lead author, Jovan Blanuša, these algorithms could also detect other illegal activities, such as phishing or manipulating stock prices. The software in question is now freely available.

J. Blanuša et al.: Graph Feature Preprocessor: Real-time Subgraphbased Feature Extraction for Financial Crime Detection. ACM International Conference on AI in Finance (2024)