<p>The integration of connected devices into the Medical Internet of Things (MIoT) has improved healthcare delivery but also brings vulnerabilities to cyber threats to patient safety and the integrity of critical medical systems. Today’s security threats are no longer simple and traditional security mechanisms are no longer sufficient. The contribution of this research is an advanced data analysis approach based on a cognitive cyber-physical system to detect and prevent cyber-attacks in MIoT environments. In this paper, we introduce a framework that combines the whale optimization algorithm (WOA) with deep learning models such as gated recurrent units (GRU) and a dense neural network (DNN) for anomaly detection in cyber-attacks. Using a set of wearable health devices and medical imaging equipment, the system is trained to detect threats with unprecedented accuracy. The performance of the model is improved with Hyperparameter optimization using WOA. Experimental results show that the proposed GRU-DNN-WOA framework achieves a detection accuracy of 98.2%, precision of 97.1%, recall of 98.0%, and F1-score of 97.5% on the MedBIoT dataset. On the IoT-23 dataset, it achieves an accuracy of 96.8%, precision of 95.7%, recall of 96.6%, and F1-score of 96.1%, outperforming prior cybersecurity techniques. The results confirm the framework's robustness, scalability, and real-time applicability for securing the MIoT systems.</p>