Binary and Multiclass Intrusion Detection in IoMT Networks: Comparative Machine Learning with McNemar Testing and XAI

The significant contribution of this research is the comprehensive evaluation of five machine-learning models for binary and 15-class IoMT intrusion detection, complemented by McNemar statistical testing and LIME-based explainability. The study demonstrates the superior performance of Decision Tree while highlighting the hidden weaknesses of minority-class detection through macro-averaged metrics.