Federated Learning for Credit Card Fraud Detection: Key Fundamentals and Emerging Trends

Abstract

With technological advancement and the evolving economic growth in modern society, acts of fraud have grown almost dramatically in the financial sector. Cybercriminals are continuously innovating sophisticated techniques to conduct illegal activities. Recent statistics underscore the severity of the issue, indicating an increase in global credit card fraud losses from ${\$}$9.84 billion in 2011 to ${\$}$32 billion in 2021, impacting various stakeholders, including individuals, businesses, and financial institutions. Machine Learning has emerged as an innovative and pivotal tool in mitigating credit card fraud by providing advanced analytical techniques for fraud detection. Nonetheless, concerns regarding data privacy and security have prompted the exploration of alternative approaches exploiting the sensitive nature of financial data. This paper systematically reviews and objectively analyzes the Federated Learning (FL) approach as an alternative to traditional machine learning by enabling collaborative model training across decentralized data sources while preserving data privacy. We explore FL’s future directions and challenges that it may encounter in credit card fraud detection and comprehensively analyze FL’s current work and outlook. We conclude by presenting several recommendations for relevant future work. This study is a valuable resource for researchers and practitioners exploring FL’s future open directions and challenges for credit card fraud detection.