An Intrusion Detection for Internet of Medical Things Based on Deep Learning Techniques

Abstract

The Internet of Medical Things (IoMT) has enhanced healthcare, but it is also vulnerable to cyber-attack. Reliable Intrusion Detection Systems (IDSs) are essential for data integrity and patient safety. The goal of this study is to design and evaluate deep learning IDSs in the form of DNNs and a hybrid GRU-DNN model for IoMT networks using the CICIoMT2024 dataset.

The design of a standalone DNN model and a hybrid Gated Recurrent Unit–DNN (GRU-DNN) was also applied and compared. Both were trained and tested with various classes involving 2, 6, and 19 from the CICIoMT2024 datasets. The GRU-DNN model, in 19-class classification performance, achieved satisfactory results of accuracy, precision, and recall as 98.4%, 98.6%,98.4% and 98.2% respectively, for F1-score. The DNN model achieved 93.1 % accuracy for the same task. The model outperformed other models, such as LSTM and previous DNN models.

The proposed hybrid GRU-DNN model exhibits promising results in being applicable to identifying intrusions in IoMT systems and seems to hold great promise for improving the security of real clinical networks