Enhancing IDS for the IoMT based on advanced features selection and deep learning methods to increase the model...

Area of Science:

• Computer Science
• Biomedical Engineering
• Cybersecurity

Background:

• Information technology, particularly the Internet of Things (IoT) and its medical variant (IoMT), is revolutionizing healthcare through remote patient monitoring.
• The increasing interconnectivity of IoMT devices presents significant security challenges, necessitating robust Intrusion Detection Systems (IDS).

Purpose of the Study:

• To propose and evaluate a novel IDS model for IoMT environments.
• To enhance the detection performance and efficiency of security systems in IoMT networks.

Main Methods:

• The proposed IDS model integrates Information Gain (IG) and Recursive Feature Elimination (RFE) for feature selection.
• A deep autoencoder (DAE) is employed for dimensionality reduction, preserving crucial data.
• A deep neural network (DNN) is utilized for classifying network traffic as normal or anomalous.

Main Results:

• The model achieved high accuracy (99.93% on WUSTL-EHMS-2020, 99.61% on CICIDS2017) and improved performance metrics (precision, recall, F1 score).
• The strategy demonstrated reduced training time and computational complexity.
• Statistical validation confirmed stable performance improvement with a p-value of 0.0001.

Conclusions:

• The developed IDS model offers superior accuracy and efficiency for IoMT security.
• The approach is effective in resource-constrained IoMT environments, enhancing overall system security.
• This study highlights the potential of advanced feature selection and deep learning for securing connected healthcare systems.