Enhancing security of internet of medical things in fog-edge environment: a study on attack detection performance

The Internet of Medical Things (IoMT) is increasingly utilized for patient health monitoring, treatment delivery, and healthcare enhancement. However, IoMT devices are vulnerable to attacks, posing risks to patient privacy and safety. Machine learning (ML) techniques are employed for attack detection to mitigate threats. IoMT devices generate substantial data, allowing ML algorithms to analyze and predict potential attacks. This study evaluates the performance of ML models for attack detection using datasets collected from medical devices. Both conventional and ensemble models are employed and trained on two datasets: one comprising all features and another with features selected based on importance. Ensemble models, which combine the strengths of multiple ML algorithms, are utilized to enhance detection accuracy and robustness. Results indicate that the ensemble model achieves superior detection rates and lower false positives than traditional ML algorithms. The stack ensemble model demonstrates the highest performance with a detection rate of 97.58%. Our proposed model outperforms existing models with a 5% improvement in detection rate and a 25% reduction in execution time. Proposed ensemble approaches hold promise for strengthening the security of IoMT devices and ensuring patient safety in fog-edge environments.