Logical Clocks Introduces new Machine Learning Technique to Detect Fraud and Prevent Money Laundering

September 10, 2020

STOCKHOLM, SWEDEN - Logical Clocks introduces a new machine learning technique to train models for fraud detection using deep learning and Generative Adversarial Networks (GANs). The technique, available on Hopsworks, the world’s first data platform with a Feature Store, helped Swedbank, the oldest and largest bank in Sweden, reduce costs associated with fighting fraud.

Fraud is a major challenge across the globe. The estimated aggregate lost turnover for companies, governments and individuals as a result of financial crimes is valued at €40 trillion, according to a report revealing the true cost of financial crime. Although there are no public numbers on the cost of financial crime in Sweden, it is estimated by local governmental organisations that the amount being laundered each year amounts to billions.

Leading financial institutions are now investing large amounts of resources in machine learning (ML) to detect and prevent fraud. However, a major challenge in applying ML for fraud detection are the constant changes such as new money transfer schemes being introduced into the market, countries joining and leaving common payment areas or currencies, and new money laundering schemes frequently appearing.

To overcome these challenges and quickly update machine learning models, Logical Clocks offers a new approach to train models for fraud detection using a semi-supervised deep learning technique known as GANs.

"GANs are now the state-of-the-art approach to combating fraud and anti-money laundering because they can overcome the class imbalance problem. That is, there are lots of examples of non-fraudulent transactions, but few examples of fraudulent transactions”, comments Dr. Jim Dowling, CEO at Logical Clocks and Associate Professor at KTH Royal Institute of Technology in Sweden.

Swedbank deployed Hopsworks technology to increase the detection rate and reduce costs of transactions associated with fraud. The bank previously employed a traditional ML rule-based system that generated 99 false-positives (alerts where the transaction did not involve fraud) for every 100 alerts. This increases the time and costs to process the data since every positive flag requires a manual investigation. After deploying Hopsworks’ GANs technique, Swedbank was able to reduce this to only 1 false-positive for every 2 alerts.

“The benefits go beyond just cost savings. The new approach makes it harder for fraudsters to avoid detection. They will no longer be able to make small adjustments in how they launder their money to get around a relatively static set of rules”, concludes Dowling.
No items found.

© Hopsworks 2024. All rights reserved. Various trademarks held by their respective owners.

Privacy Policy
Cookie Policy
Terms and Conditions