Intrusion Detection and Real-Time Adaptive Security in Medical IoT Using a Cyber-Physical System Design.

Area of Science:

• Cybersecurity
• Medical Informatics
• Artificial Intelligence

Background:

• Medical Internet of Things (MIoT) devices are increasingly vital in healthcare, but present significant cybersecurity risks.
• Cyber incidents targeting MIoT systems pose direct threats to patient safety and data integrity.
• Existing defense mechanisms struggle to adapt to the evolving threat landscape in MIoT environments.

Purpose of the Study:

• To introduce a novel machine learning-enabled Cognitive Cyber-Physical System (ML-CCPS) for enhanced cybersecurity in MIoT.
• To develop an adaptive defense mechanism capable of identifying and responding to sophisticated cyber threats in real-time.
• To improve the security and reliability of life-critical healthcare systems reliant on connected devices.

Main Methods:

• Development of a layered cognitive architecture for the ML-CCPS.
• Integration of hybrid feature modeling and physical behavioral analysis.
• Utilization of Extreme Learning Machine (ELM) for adaptive access control, continuous monitoring, and intrusion detection.

Main Results:

• The ML-CCPS achieved a macro F1-score of 97.8% and an AUC of 99.1% on the ToN-IoT dataset.
• Demonstrated robust performance under noisy telemetry and in resource-constrained environments.
• Effectively detected novel attack types and scaled with an increasing number of connected devices.

Conclusions:

• The ML-CCPS offers a highly effective and reliable solution for securing MIoT environments against cyber threats.
• The system provides adaptive access control and intrusion detection with acceptable computational overhead.
• Validated through comparative evaluations and ablation studies, the ML-CCPS is suitable for real-time MIoT security applications.