Scholarlink

<p>IoHT serves a vital&#xa0;role in everyday life by&#xa0;providing&#xa0;electronic health services and improving&#xa0;the treatment of patients. In recent years, IoHT devices and applications have experienced rapid growth, making them increasingly vulnerable to cyberattacks. Thus, IoHT devices need to be safeguarded to preserve electronic health information. To solve the issue, the proposed method implements a hybrid many-objective evolutionary optimization and ML algorithm-based feature selection approach for a network intrusion detection system (NIDS) in the IoHT environment. The study simultaneously optimizes (i.e., maximizes) multiple performance metrics (recall, specificity, F1-score, accuracy, precision). First, data preprocessing is performed over the dataset using feature encoding and feature scaling. For feature selection, the proposed technique utilizes four many-objective evolutionary algorithms (NSGA-III, AGE-MOEA, MOEA/D, SMS-EMOA) and one ML classifier (Extra Trees classifier). After feature selection is performed, four ML classifiers (KNN, Extra-Trees classifier, MLP, and random forest) are utilized for attack detection in the IoHT environment. The comparative analysis shows that the proposed strategy outperforms the existing IDS methods with 99.63% recall, 99.65% F1-score, 99.67% precision, 99.83% accuracy, 99.89% specificity (under Bluetooth-BR/EDR dataset), and 99.40% recall, 99.63% F1-score, 99.86% precision, 99.42% accuracy, 99.48% specificity (under ECU-IoHT dataset), and 100.00% recall, 98.17% F1-score, 96.41% precision, 99.51% accuracy, 99.44% specificity (under WUSTL-EHMS dataset) for IoHT attack detection. Among the four metaheuristic optimizations, SMS-EMOA performed better than other state-of-the-art metaheuristic algorithms (NSGA-III, AGE-MOEA, MOEA/D) in all three benchmark datasets for network-based intrusion detection under IoHT networks.</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

A hybrid MaOEA-ML framework for network intrusion detection system in the Internet of Health Things (IoHT)

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • Shashank Sawan,
  • B Ramachandra Reddy

摘要

IoHT serves a vital role in everyday life by providing electronic health services and improving the treatment of patients. In recent years, IoHT devices and applications have experienced rapid growth, making them increasingly vulnerable to cyberattacks. Thus, IoHT devices need to be safeguarded to preserve electronic health information. To solve the issue, the proposed method implements a hybrid many-objective evolutionary optimization and ML algorithm-based feature selection approach for a network intrusion detection system (NIDS) in the IoHT environment. The study simultaneously optimizes (i.e., maximizes) multiple performance metrics (recall, specificity, F1-score, accuracy, precision). First, data preprocessing is performed over the dataset using feature encoding and feature scaling. For feature selection, the proposed technique utilizes four many-objective evolutionary algorithms (NSGA-III, AGE-MOEA, MOEA/D, SMS-EMOA) and one ML classifier (Extra Trees classifier). After feature selection is performed, four ML classifiers (KNN, Extra-Trees classifier, MLP, and random forest) are utilized for attack detection in the IoHT environment. The comparative analysis shows that the proposed strategy outperforms the existing IDS methods with 99.63% recall, 99.65% F1-score, 99.67% precision, 99.83% accuracy, 99.89% specificity (under Bluetooth-BR/EDR dataset), and 99.40% recall, 99.63% F1-score, 99.86% precision, 99.42% accuracy, 99.48% specificity (under ECU-IoHT dataset), and 100.00% recall, 98.17% F1-score, 96.41% precision, 99.51% accuracy, 99.44% specificity (under WUSTL-EHMS dataset) for IoHT attack detection. Among the four metaheuristic optimizations, SMS-EMOA performed better than other state-of-the-art metaheuristic algorithms (NSGA-III, AGE-MOEA, MOEA/D) in all three benchmark datasets for network-based intrusion detection under IoHT networks.