More secure payment thanks to machine learning

Internet fraudsters are becoming more and more sophisticated – smart protective measures are required. The fraud detection solution developed by Ergon Informatik AG can detect fraudulent payments more reliably than traditional systems, thanks to artificial intelligence.

The fraud detection solution is a self-learning system that analyses huge amounts of data from payment transactions, independently identifies regularities and detects irregularities and fraud attempts quickly and with high accuracy.

Two years ago, the software company Ergon Informatik from Zurich, which specialises in financial services and industry, set up a pilot project in which machine learning was tested for better online security. The basis was the payment data of a major financial institution. The performance of artificial intelligence was compared to traditional methods where human experts make the decision "fraud – yes or no". The result: even in the first round, the newly developed fraud detection solution identified cases of fraud more reliably than human experts. The number of false positive cases, i.e. transactions that are falsely considered fraudulent, was also lower. The system is now in use at several banks and self-learning is further increasing its hit rate.