A model for managing online fraud risk using transaction validation

ABSTRACT Online banking fraud is about deceiving the banking system so that it appears that a transaction is authentic when it is not. Customer experience and fraud losses are two critical aspects to consider when implementing online fraud models. Maintaining the right balance between the two extremes is driven by the bank's operational risk management philosophy. This paper discusses transaction validation of an online banking fraud detection model implemented by a US bank. The validation is focused on measuring the model's effectiveness in capturing online fraudulent transactions. The paper summarizes various model weaknesses including: the need for a justified documented procedure for setting performance metrics and tolerance limits; the calibration of key metrics to reflect the nature of online fraud behavior; and the model's inability to link transactions with the final outcome as false negatives or true negatives, which are not captured by the model. As well as these, various metrics are discussed in the paper, which can be used as a guideline when setting performance metrics for an online banking fraud detection model.