Remote Banking Fraud Detection Framework Using Sequence Learners

The reliability and performance of fraud detection techniques has been a major concern for the financial institutions as traditional fraud detection models couldn’t cope with the emerging new and innovative attacks that deceive banks. This paper proposes a conceptual fraud detection framework that can detect anomalous transaction quickly and accurately and dynamically evolve to maintain the efficiency with minimum input from subject matter expert. Based on the proposed framework,we implement Long Short-Term Memory (LSTM) based Recurrent Neural Network model for detecting fraud in remote banking and evaluate its performance against SVM models. Two novel features for remote banking fraud are evaluated, i.e., the time spent on a page and the time between page transitions. Modeling is performed on an anonymised real-life dataset, provided by a large financial institution in Europe. The results of the modeling demonstrate that given the labeled dataset both the LSTM and SVM model can detect payment fraud with acceptable accuracy, though overall the LSTM models perform slightly better than the SVM models. The results also prove the hypothesis that the events across banking channels can be modeled as time series data and then sequence-based learners such as Recurrent Neural Network (RNN) can be applied to improve or reduce the False Positive Rate (FPR) and False Negative Rate (FNR).