Digital fraud has become a critical concern across various sectors, making it essential for organizations to adopt robust detection and prevention mechanisms while reinforcing overall cyber security infrastructures. Machine learning techniques have shown strong potential in analysing high-volume transactional data to detect patterns and irregularities that signal fraudulent behaviour. In this study, a heterogeneous dataset containing both genuine and fraudulent UPI transactions is utilized to develop an intelligent fraud detection framework. The dataset undergoes systematic data pre-processing and feature engineering to extract key attributes such as transaction amount, time of transaction, payer and payee details, geographical location, and device identifiers. Special focus is given to temporal and behavioural patterns to effectively capture time-dependent fraud characteristics. The machine learning model is trained on historical data and evaluated based on classification accuracy, precision, recall, and F1-score to ensure reliability in distinguishing legitimate transactions from fraudulent ones. The optimized model is then integrated into the Unified Payments Interface (UPI) environment to enable real-time fraud detection. Furthermore, an automated alert system is incorporated to facilitate immediate notifications and intervention when suspicious activity is detected. The proposed framework enhances the security, efficiency, and trustworthiness of digital payment systems by proactively addressing and mitigating fraud risks.
JALIB et al. (Mon,) studied this question.