Scholarlink

<p>This paper presents a resource-efficient IDS tailored for the IoMT, primarily emphasizing IoMT-enhanced security in medical networks. Characteristically resource-constrained and inherently vulnerable to various cyber threats, IoMT devices are typically ordinary. Inherent IoMT IDS pays off with the necessity for a critical trade-off between detection accuracy and computational efficiency. For this reason, the two-step feature selection stage is proposed to bring this challenge down. Firstly, the system decreases the dimensionality by operating the MI filtering on the dataset and keeps only the most informative features. It further refines this using ensemble-based ranking methods, such as Random Forest, AdaBoost, XGBoost, and LightGBM, to ensure the optimum feature selection for the task. Random Forest was adopted for the final classification because it is generally robust, efficient at handling high-dimensional data, and usually performs well. The proposed system has been tested with intensive usage using three well-acknowledged benchmark datasets, namely WUSTL-EHMS-2020, NSL-KDD, and CIC-IoMT2024. It showed considerable accuracy, precision, recall, and F1-score gains, particularly for DDoS and DoS attack types. The proposed technique significantly reduces execution time and memory usage. This, in turn, makes the approach much more suitable for real-time implementation. At the same time, the gains to be had in computational resource savings make this an attractive approach for resource-constrained hardware environments. This work proposes an effective and efficient IDS for IoMT security.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An optimized intrusion detection system for resource-constrained IoMT environments: enhancing security through efficient feature selection and classification

  • 期刊论文

  • 发表日期：

    2025年4月27日

查看全文
  • Arash Salehpour,
  • Mohammad Ali Balafar,
  • Alireza Souri

摘要

This paper presents a resource-efficient IDS tailored for the IoMT, primarily emphasizing IoMT-enhanced security in medical networks. Characteristically resource-constrained and inherently vulnerable to various cyber threats, IoMT devices are typically ordinary. Inherent IoMT IDS pays off with the necessity for a critical trade-off between detection accuracy and computational efficiency. For this reason, the two-step feature selection stage is proposed to bring this challenge down. Firstly, the system decreases the dimensionality by operating the MI filtering on the dataset and keeps only the most informative features. It further refines this using ensemble-based ranking methods, such as Random Forest, AdaBoost, XGBoost, and LightGBM, to ensure the optimum feature selection for the task. Random Forest was adopted for the final classification because it is generally robust, efficient at handling high-dimensional data, and usually performs well. The proposed system has been tested with intensive usage using three well-acknowledged benchmark datasets, namely WUSTL-EHMS-2020, NSL-KDD, and CIC-IoMT2024. It showed considerable accuracy, precision, recall, and F1-score gains, particularly for DDoS and DoS attack types. The proposed technique significantly reduces execution time and memory usage. This, in turn, makes the approach much more suitable for real-time implementation. At the same time, the gains to be had in computational resource savings make this an attractive approach for resource-constrained hardware environments. This work proposes an effective and efficient IDS for IoMT security.