Most fraud management systems identify potential frauds by sifting through data to identify patterns which violate some pre-configured rules. The potential weakness is obvious: if the rules are not set correctly, fraud is not identified. On the other hand, rules may also be ineffective if they generate too many false positives, wasting the time of human analysts who have to disregard them. Rules also need to be modified over time, to reflect changes in circumstances and adapt to new types of behavior adopted by fraudsters. Fraudsters learn what frauds they can get away with by trial and error; when they are caught, they may modify their approach and try again, so it is vital to evolve to keep pace with fraudsters. Life would be easier if computers could emulate the fraudsters, learning about fraud from experience, without needing to be programmed with lots of rules that then need to be maintained. So can they?
Shankar Palaniandy is one businessman and fraud expert who believes software is more effective if it designed to learn, instead of following rules. He established FRS Labs and patented a Fraud Management System (FMS) that searches for patterns of fraudulent behavior without being tied to rules. And he has sold this system to several opcos in the Vodafone Group. So what is Shankar’s secret? How did he persuade Vodafone India to ditch the rules and use his FMS instead? Why is his business able to challenge larger and more established software vendors? We invited Shankar to join us for episode 23 of the Commsrisk podcast, and he gave us some fascinating insights into the development of a different kind of FMS, and why it has gained traction in the market place.
My co-host, Dan Baker, also contributed his insights during the podcast. Dan has been doing a lot of research into FMS recently, having just published his new report on anti-fraud solutions and strategies.
To hear Shankar’s interview with Dan and I, you can press the ‘play’ button on this page, or you can download the mp3 file from here. You can also obtain this podcast, and all future podcasts, by subscribing via our page at the iTunes Store, or our page at the Blubrry Android subscription service.