Enhancing Federated Learning Security with Reputation-Based Phishing Defense

This study extends the innovative research presented in “FlPhish: Reputation-Based Phishing Byzantine Defense in Ensemble Federated Learning,” IEEE ISCC 2021. It tackles the critical issue of securing Federated Learning (FL) systems from Byzantine adversaries capable of sabotaging model performance through malicious client behavior. The proposed architecture, Ensemble Federated Learning (EFL), incorporates a groundbreaking phishing mechanism and a Bayesian-based reputation system to identify and mitigate such threats effectively. This approach integrates a weighted aggregation method that enhances the resilience of EFL against malicious actors.