Fraud ain’t what it used to be

Between the time this blog was published and you reading it, new fraud schemes will have emerged. Indeed, with the ever increasing number of, and methods for, payment transactions and the new supporting technologies, not least mobile and internet banking, the new opportunities for fraudsters are increasing.

Banks are realising that the trends in payments are for faster, cheaper mechanisms that demand increased complexity not just to deliver the payment but also the appropriate fraud, risk and credit checks. These become exponentially more challenging to do effectively as the time between initiation and irrevocable settlement of the transaction decreases, and the range of payment channels increases.

In terms of fraud detection, typically a layered approach is used with a mix of tools and methods to provide multiple layers of defence. Traditional methods, including predictive models are proven for established fraud patterns, but are not agile enough to react immediately as new methods of fraud emerge. It takes time for these traditional systems to accumulate sufficient data and train the models to detect the emerging fraud, leaving a window of opportunity for the fraudster until these systems catch up.

This reaction speed element is increasingly important as the fraudsters find a gap that they can exploit they will use technology to seek to maximise their return and will often share new practices with other fraud rings – hence new schemes now propagate like wildfire. Minimising the amount of time that they have to exploit any gap will significantly reduce the risk, disruption and reputational damage for the financial institution.

To counteract these rapid emerging schemes, firms are looking to use what are called “flash fraud” models. In the fraud arms race, the ability to rapidly implement flash fraud models is the equivalent of an army’s rapid reaction force that is deployed to stop unexpected invasions. These flash fraud models require a rapid combination of recent, relevant sets of transaction data and the expertise of fraud experts using tools and techniques to boost the fraud signal.

Together these can provide a set of updated and more effective rules in a matter of hours rather than days or weeks. For example, a major payment processing company dramatically improved its response to flash fraud, reducing the time to deploy new fraud models from two weeks to a single day. Using their flash fraud solution, analysts review recent transactions every day, analyse them to uncover potential patterns, generate business rules to flag those patterns, and then, implement the rules into their payments system.

This is an exciting and challenging area we are working in, and for those institutions that can resolve these challenges there are major benefits to be realised.