AI-Driven Anomaly Detection and Facial Recognition for Financial Cybersecurity: A Hybrid Approach to Fraud Prevention

Banks are facing increasing cyberattacks, including fraud and insider threats. Traditional rule-based anomaly detection is not effective against sophisticated fraud techniques. In this paper, a hybrid AI-based anomaly detection system is proposed that combines supervised and unsupervised learning models to improve detection accuracy. Real-time log monitoring, behavior analysis, and facial recognition are employed to improve fraud prevention. Isolation Forests, Autoencoders, and XGBoost are utilized for anomaly detection and provide compliance support for financial regulations like GDPR and PCI-DSS. Experimental results confirm its effectiveness in reducing false positives and real-time fraud detection.