An Explainable Differential Privacy Framework for IoMT Using Adaptive Gaussian Noise

The rapid growth of the Internet of Medical Things (IoMT) has increased privacy and security concerns for sensitive healthcare data. This research proposes Hybrid Separate Dynamic Gaussian Noise Distribution with Coyote Badger Optimization (HSDGN-CBO), a privacy-preserving framework combining Flipped Huber Distribution (FHD) with Coyote Badger Optimization (CBO) to improve the privacy-utility balance by reducing noise while maintaining strong privacy guarantees. An edge-assisted federated learning framework with federated intrusion detection enables secure collaborative learning without sharing raw medical data. Experiments on Breast Cancer Prediction and COVID-19 Home Nursing datasets achieved a 0.95 privacy score, 3 kWh energy consumption, and 72-iteration convergence, demonstrating a scalable, efficient, and interpretable IoMT privacy solution.