A Survey :Federated Learning Driven Decentralized Security In Digital Payment Systems

Authors: Professor Shylaja B, Panuganti Snigdha, Reekanksha Prakash, Repakula Tharuni, Rithika Shankar

Abstract: The financial sector is undergoing rapid digital transformation, accompanied by a surge in cyber threats and fraud. Traditional centralized machine learning approaches for fraud detection are increasingly limited by privacy concerns, data-sharing restrictions, and regulatory compliance issues. Federated Learning (FL) offers a decentralized alternative by enabling collaborative model training across institutions without sharing sensitive data. This survey explores the application of FL in financial security, focusing on its foundations, privacy-preserving mechanisms, and real-world use cases such as fraud detection, credit scoring, and customer behavior analysis. We compare FL with existing centralized techniques in terms of accuracy, privacy, adaptability, and scalability. Additionally, we examine how FL integrates with emerging technologies like blockchain, Explainable AI (XAI), and Secure Multi-Party Computation (SMPC). The paper highlights key challenges, research gaps, and future directions, providing a comprehensive overview of FL's potential to revolutionize secure and intelligent financial systems.