TOMLDIoMT-A Trust based Optimized Machine learning Detector for Internet of Medical Things

ABSTRACT The Internet of Medical Things (IoMT), an extension of the Internet of Things (IoT), enables continuous health monitoring through interconnected medical sensors. Despite its benefits, IoMT networks are highly vulnerable to network-based attacks such as spoofing and data manipulation, where malicious nodes attempt to mislead neighboring devices. Ensuring secure communication among trustworthy nodes is therefore critical for reliable IoMT deployment. This paper proposes a Machine Learning–based Trust Management System (TMS) to detect and prevent malicious nodes in IoMT networks. The proposed framework computes real-time trust values using a novel feature set that integrates Direct Trust (behavior-based), Indirect Trust (reward–punishment-based), and Data Trust (sensor deviation-based). The model is evaluated on two benchmark datasets, WUSTL2020 and ROUT-4-2023. Challenges such as class imbalance and missing values are addressed using SMOTE and preprocessing techniques. Multiple machine learning classifiers—including ELM, XGBoost, Random Forest, KNN, SGB, and Logistic Regression—are trained. The best hyperparameters of each model is selected using an automated optimization framework OPTUNA. Experimental results demonstrate that the Random Forest model achieves superior performance, with accuracy ranging from 99.7% to 99.93% and a False Alarm Rate between 0% and 0.1%. The proposed framework outperforms existing state-of-the-art trust management solutions for IoMT networks.