Anomaly-Based Intrusion Detection System for the Internet of Medical Things

Abstract

The use of the Internet of Things (IoT) in the health sector, known as the Internet of Medical Things (IoMT), allows for personalized and convenient (e)-health services for patients. However, there are concerns about security and privacy as unethical hackers can compromise these network systems with malware. We proposed using hyperparameter-optimized Machine and Deep Learning models to address these concerns to build more robust security solutions. We used a representative Anomaly Intrusion Detection System (AIDS) dataset to train six state-of-the-art Machine Learning (ML) and Deep Learning (DL) architectures, with the Synthetic Minority Oversampling Technique (SMOTE) algorithm used to handle class imbalance in the training dataset. Our hyperparameter optimization using the Random search algorithm accurately classified normal cases for all six models, with Random Forest (RF) and K-Nearest Neighbors (KNN) performing the best in accuracy. The attention-based Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) model was the second-best performer, while the hybrid CNN-LSTM model performed the worst. However, there was no single best model in classifying all attack labels, as each model performed differently in terms of different metrics.