Real Time Fraud Detection System for Mobile Banking: Based on Experiential Paradigm

Abstract

The current banking industry is characterized by hyper-competition driven by technological innovations that revolve around provision of ubiquitous access to banking services especially through mobile banking. Proliferation of mobile phones in Kenya acts as a substrate for the increased adoption of mobile banking in Kenya. Frauds perpetrated through mobile banking platforms have become prevalent eroding the hard-earned profits by banks. This research therefore was aimed at developing a case-based reasoning framework that would do real time fraud detection in mobile banking. Case-based reasoning problem solving technique which makes use of prior knowledge and specific problem scenarios (cases) to solve new problems by identifying similar past problem episodes and applying them to the new problem situations. The research employed an incremental prototyping model in which the overall architectural design was done upfront but the detailed design and developments of the subcomponents were done in incremental manner. The research used a four-step approach for building the Case Based Reasoning engine which included features calibration, case stabilization, and implementation and finally the evaluation process. The research relied on both primary and secondary data to collect the past fraud incidences to build a reference case library. The research design was in form of interviews done to the target population comprising of individuals drawn from the bank’s risk, forensics, digital channels support and information systems security. The Case Based Reasoning algorithm implemented incorporated a threshold retrieval mechanism combined with K-Nearest Neighbor algorithm. The system prototype was built and trained using a data set of 120 transactions with system evaluation done in three iterations of 40 transactions in every iteration revealing an average classification accuracy of 84.17%.