A Secure and Explainable Federated Intrusion Detection System Using Deep Learning and Metaheuristic Optimization for Healthcare IoT

This paper introduces a new AI-based threat detection model optimized for healthcare infrastructures which include IoT-integrated monitoring systems and implementing multi-cloud services. The developed system is based on and makes use of hybrid deep learning algorithms, homomorphic encryption, and zero trust security to achieve real time cyber threat detection and maintain patient privacy. Experimental environment A model healthcare environment was simulated with virtual machines, cloud emulators, and edge IoT devices in order to evaluate the system performance. We evaluated the model on both synthetic attacks datasets and real-time IoT feeds, comparing it with respect to detection accuracy, false positive rate, latency, throughput, and privacy overhead. The experimental results show the performance is better than the traditional intrusion detection system, the detection accuracy is 98.7%, the latency is low (42ms), and the rate of false positives is lower which is 1.2%. Microsoft SEAL for encrypted analytics and Keycloak for role-based access control is combined in GMLOS to ensure the data confidentiality. This paper establishes the possibility and effectiveness of a secure, scalable and intelligent IDS infrastructure designed for the next generation of healthcare systems.