The latest developments in the Internet of Medical Things (IoMT) have transformed healthcare, allowing for the easy transmission of sensitive data across networked medical equipment. However, this interconnection has produced a fertile environment for cybersecurity threats, necessitating the use of sophisticated intrusion detection systems. This work proposes a fresh classification of an attack detection algorithm for IoMT, as well as a method comparison and dataset categorization to improve detection performance. The aim of research is to propose and evaluate an attack detection algorithm for IoMT, focusing on improving detection performance against man-in-the-middle attacks. The study focuses on man-in-the-middle attacks, a common concern for the security of IoMT communication. The suggested intrusion detection technique employs machine learning algorithms, where real-time data from the St. Louis Enhanced Healthcare Monitoring System (WUSTL-EHMS) is used for evaluation, proving the Variational Autoencoders (VAEs) better performance with 91.61% accuracy. The study not only offers information on the usefulness of deep learning in cyber-attack detection, but it also indicates the changing environment of IoMT cybersecurity concerns. The investigation makes use of an eight variety of machine learning models, including feedforward neural networks, gradient boosting machines (XGBoost and LightGBM), logistic regression, VAEs, random forests, along with support vector machines (SVM). These models are assessed using measures such as accuracy, precision, recall, and F1 score. The findings define the significance of adjusting to the changing nature of cyber threats in IoMT, with deep learning algorithms playing a critical role in strengthening security measures. Integrating AI-driven intrusion detection systems is crucial for safeguarding IoMT against evolving cyber threats, ensuring patient safety and data privacy.
