This document introduces SRF2T-ID, a two-tiered intrusion detection model designed to enhance security and privacy in smart residency environments. The inner layer focuses on securing individual smart homes through techniques like AES encryption, packet division, clock synchronization, and randomization of transmission channels and receiver pipes. The outer layer implements a federated learning-based Network Intrusion Detection System (NIDS), employing a binary-multiclass classification approach and nature-inspired meta-heuristic feature selection algorithms to protect the overall network. The research utilizes CICIDS2017 and CICIoT2023 datasets for evaluation, demonstrating high detection accuracy, and includes a real-time test-bed implementation to prove its practical viability.