Tackling online fraud

Online fraud is an unnecessary headache for online merchants. According to CyberSource Corp. online merchants were estimated to lose $3.0 billion from fraud. Although this is a small percentage of revenue which in 2006 was 1.4%, it is the manual review process that hurts the most. CyberSource estimates that 81% of merchants are manually checking orders. The average rate of manual review exceeds 1 out of 4 orders.

How should business tackle online fraud?

Typically, depending on the size of the business, a three-pronged strategy is applied to tackle fraud, namely - (a) rule-based (2) neural network algorithms (3) data mining techniques. The challenges of the available technologies are (1) all the techniques are based on historical data (2) they are unable to detect fraud as they occur (3) neural network and data mining techniques are hard to implement and are expensive.

One of the best ways of tackling online fraud is to catch the fraudulent transaction as the transactions happen – in real-time. Businesses have a variety of techniques and software at their disposal to implement real-time fraud check. However, since such rules are CPU intensive, they increase the latency of processing the transaction. More rules lead to non-linear increase in processing time. At the same time, the budget to process a payment transaction cannot exceed the limits set by the Networks (such as Visa and MasterCard).

Hence what business needs is a low-latency payment processing software that can authorize transactions as well as process multiple rule-based fraud techniques (such as Address Verification Service, Card Verification Number, Blacklist check, Velocity monitoring and more) in real-time without impacting the time budget for authorizing a transaction. The good news is that such payment software is available today based on next-generation application server farmework that authorize thousands of transactions per second in single digit millisecond latency.