Towards Effective Classification of Intrusion in Internet of Medical Things using Random Forest Classifier for Sustainable Health Care

Information Technology is extensively utilized in contemporary healthcare settings. Nevertheless, healthcare facilities are frequently targeted by malicious actors aiming to disrupt services and exfiltrate sensitive data. Intrusion detection systems play a crucial role in monitoring and analyzing computer systems or networks for indications of such attacks. To safeguard healthcare technologies, including the Internet of Medical Things (IoMT), it is imperative to protect them from these threats. Machine learning models have demonstrated superiority over traditional methods in detecting intrusions. However, some of these models still exhibit limitations, such as generating false alarms when identifying attacks on IoMT. Consequently, the development of more accurate models for detecting these attacks is essential. Security experts have developed datasets that simulate various attack scenarios for testing purposes, one of which is the IoMT Attack Testbed. This study proposes a three-step methodology for constructing an effective intrusion detection model. It employs a Random Forest classifier to categorize intrusions within the dataset. Through dataset analysis, the study provides insights into feature handling, addresses data imbalance issues, and identifies significant features for the model. The model's parameters were optimized to enhance its performance. The model was evaluated in two scenarios, with results indicating superior performance in the second scenario when data imbalance was addressed, critical features were selected, and parameters were fine-tuned.