UPI Fraud Detection

Abstract

The growing trend of digital payments in India and especially the emergence of Unified Payment Interface (UPI) is associated with a rise in the number of cases of financial fraud online. Current fraud detection techniques are based on the use of simple rule-based methods like restricting transaction amount from exceeding some threshold limit but prove to be inefficient in dealing with the increasingly sophisticated nature of the frauds committed using UPI. This paper describes the development of UPI Fraud Detection System based on the machine learning algorithm RandomForest for analysing transaction behavior patterns and categorizing transactions into three risk categories: Safe, Suspicious and High Risk. The UPI Fraud Detection System has been developed as a full-fledged web application in Python and Flask Framework. It has been trained using the real Kaggle Credit Card Fraud Detection dataset containing 2,84,807 financial transactions. The system demonstrates the feature of OTP verification process for transactions classified as high-risk ones.