IBM to tackle fraud with Iris Analytics

15.01.2016
IBM is going to apply machine learning to fraud busting with Iris Analytics.

While that makes it sound as though it will be using Watson AI systems to identify fraudsters by gazing deep into their eyes, this is really about its acquisition of a German software firm called Iris Analytics.

Iris monitors banking transactions and uses machine learning to spot previously unknown patterns of fraudulent transactions in real time. The system can work alone or in conjunction with human analysts, according to IBM

With only one bank in six equipped with real-time fraud detection systems, and even those taking a month or more to learn to stop new attacks once they are identified, IBM sees a big market for integrating systems like that of Iris with its existing antifraud products.

This is far from IBM's first move into the antifraud market.

In September 2013, it bought Israeli antifraud specialist Trusteer, which specialized in monitoring financial services and provided a browser protection plugin for online banking.

Early the following year it rolled out a real-time fraud detection tools, Counter Fraud Management Software, building on its previous acquisitions of Cognos, SPPS, i2 and FileNet.

IBM hopes that the new "cognitive computing" approach taken by Iris will allow it to speed up and scale up its fraud detection systems -- while also allowing customers to respond more rapidly by lowering the number of false positives that must be investigated.

The acquisition will bring it a number of existing Iris clients in the payment and banking industry, including the French Interbank payment card processing network e-rsb operated by Stet. According to IBM, Iris adds around 5 milliseconds to the processing time for e-rsb transactions. That's a small addition to the hundreds of milliseconds the network takes to process each of the 750 transactions it handles per second, according to the e-rsb website.

It's not just big banks or their payment processors that are looking to machine learning to reduce fraud: smaller banks are buying such services too. For example, the Orrstown community bank in Pennsylvania built its fraud detection system using Splunk and the Prelert behavioral analytics package.

With the introduction of Chip and PIN shifting the liability for fraud, and future payment systems likely to cause similar disruptions, its more important than ever to make a quick decision about whether a payment is suspicious or safe to accept, according to Iris.

Peter Sayer