Scalable Federated Learning for Privacy-Preserving Credit Card Fraud Detection

In the modern digital landscape, credit card fraud remains a persistent threat, posing significant challenges to both individuals and banks, leading to privacy breaches. This study introduces an innovative approach by integrating federated learning into credit card fraud detection. Unlike traditional methods relying on centralized data collection, federated learning allows several banks to work together in training a fraud detection model while safeguarding individual transaction privacy through secure aggregation and differential privacy techniques. Experimental results with real-world data show that this approach achieves comparable performance to centralized methods, offering an effective and privacy-preserving solution to combat credit card fraud, enhance financial security, and protect sensitive data. This research provides a practical and secure means for banks to collectively identify credit card fraud and provide data privacy for both banks and cardholders.