FBID Framework for Robust OOD Attack Detection in IoT

FBID uses a contextual multi-armed bandit at the server to regulate client-side training intensity in personalized federated learning. This prevents over-personalization in non-IID IoT environments, using a trust-based blending mechanism to interpolate between global and local models for better out-of-distribution attack detection.