Financial transactions have been transformed by the rapid development of digital payment systems such the Unified Payments Interface (UPI). However, this growth has also contributed to a rise in payment fraud across a number of platforms, including mobile payment systems, e-commerce and credit cards as traditional fraud prevention methods are increasingly becoming inadequate. This therefore calls for more robust detection mechanisms such as Machine Learning (ML) and biometric authentication, to prevent expert fraudsters to continue preying on vulnerabilities and inflicting huge financial losses around the world. This study explores the use of advanced Machine Learning (ML) techniques in conjunction with contemporary biometric identification systems to further identify and stop fraudulent Unified Payments Interface (UPI) transactions, taking into account the limitations of traditional fraud detection techniques. We explore the possibilities of fingerprint scanning, facial recognition, behavioural biometrics and other cutting-edge approaches to enhancing UPI security in addition to traditional security measures. Using a Kaggle-sourced UPI dataset (50,000 records), we preprocess data (handling missing values, normalization, feature selection) and train ML models with an accuracy of 95.3% accuracy (Random Forest). When fused with biometrics, detection is boosted to 99.1%, resulting to an improvement by 3.8% over individual ML. Power BI enables real-time anomaly visualization. Our work bridges critical gaps in UPI-specific fraud research and demonstrates the necessity of multi-layered security for dynamic payment systems.