Abstract

The Unified Payment Interface (UPI) has transformed the landscape of digital payments, significantly enhancing the pace of digitization. It has redefined the financial ecosystem by facilitating quicker, simpler, and more accessible cashless transactions for individuals across the nation. However, the increasing use of UPI has also made it a target for various fraudulent activities. Therefore, it is essential to detect fraudulent transactions in real-time by leveraging artificial intelligence techniques. Many machine learning algorithms are used to detect fraudulent transactions, but most of them rely heavily on historical transactional data to identify patterns and anomalies. While historical data provides valuable insights into past fraud patterns, it may not be sufficient to detect dynamic patterns in evolving fraudulent transactions. We propose a CNN-based fraud detection model that recognizes complex patterns in real-time transactional data. In this paper, we analyze the performance of machine learning algorithms and a CNN-based approach. Our results demonstrate that the CNN-based model outperforms traditional machine learning models in detecting UPI frauds.