Fraud Detection in UPI Transaction Using Machine Learning

The use of UPI has grown significantly in recent years. Fraud instances linked to UPI are increasing as it becomes the most widely used payment method for both regular and online purchases. In this research, we use a Convolutional Neural Network (CNN) to model the steps involved in processing UPI transactions and demonstrate how fraud detection may be done with it. Initially, a CNN is trained using a cardholder's typical behaviour. An inbound UPI transaction is deemed fraudulent if the trained CNN does not accept it with a high enough likelihood. At the same time, we work to prevent the rejection of legitimate transactions. To demonstrate the efficacy of our strategy and to contrast it with other methods found in the literature, we provide comprehensive experimental results.