Optimizing Intrusion Detection in IoMT Networks Through Interpretable and Cost-Aware Machine Learning DOI Creative Commons
Abdelatif Hafid, Mohamed Rahouti, Mohammed Aledhari

et al.

Mathematics, Journal Year: 2025, Volume and Issue: 13(10), P. 1574 - 1574

Published: May 10, 2025

The rise of the Internet Medical Things (IoMT) has enhanced healthcare delivery but also exposed critical cybersecurity vulnerabilities. Detecting attacks in such environments demands accurate, interpretable, and cost-efficient models. This paper addresses challenges network security, particularly IoMT, through advanced machine learning (ML) approaches. We propose a high-performance framework leveraging carefully fine-tuned XGBoost classifier to detect malicious with superior predictive accuracy while maintaining interpretability. Our comprehensive evaluation compares proposed model well-regularized Logistic Regression baseline using key performance metrics. Additionally, we analyze security-cost trade-off designing ML systems for threat detection employ SHAP (SHapley Additive exPlanations) identify features driving predictions. further introduce late fusion approach based on max voting that effectively combines strengths both Results demonstrate achieves higher (0.97) recall (1.00) compared Regression, our provides more balanced improved precision (0.98) reduced false negatives, making it suitable security-sensitive applications. work contributes developing robust, efficient solutions addressing evolving networked environments.

Language: Английский

Optimizing Intrusion Detection in IoMT Networks Through Interpretable and Cost-Aware Machine Learning DOI Creative Commons
Abdelatif Hafid, Mohamed Rahouti, Mohammed Aledhari

et al.

Mathematics, Journal Year: 2025, Volume and Issue: 13(10), P. 1574 - 1574

Published: May 10, 2025

The rise of the Internet Medical Things (IoMT) has enhanced healthcare delivery but also exposed critical cybersecurity vulnerabilities. Detecting attacks in such environments demands accurate, interpretable, and cost-efficient models. This paper addresses challenges network security, particularly IoMT, through advanced machine learning (ML) approaches. We propose a high-performance framework leveraging carefully fine-tuned XGBoost classifier to detect malicious with superior predictive accuracy while maintaining interpretability. Our comprehensive evaluation compares proposed model well-regularized Logistic Regression baseline using key performance metrics. Additionally, we analyze security-cost trade-off designing ML systems for threat detection employ SHAP (SHapley Additive exPlanations) identify features driving predictions. further introduce late fusion approach based on max voting that effectively combines strengths both Results demonstrate achieves higher (0.97) recall (1.00) compared Regression, our provides more balanced improved precision (0.98) reduced false negatives, making it suitable security-sensitive applications. work contributes developing robust, efficient solutions addressing evolving networked environments.

Language: Английский

Citations

0