Scholarlink

<p>Security issues have increased due to the increasing amount of IoT devices, making good intrusion detection systems (IDS) necessary to counteract changing cyberthreats. In order to overcome these obstacles, this work suggests combining a Kernel Extreme Learning Machine (KELM) method with a Binary Emperor Penguin Optimization (BEPO) algorithm-based feature selection for effective classification. Two benchmark datasets, CIC-IDS-2018 and Bot-IoT, were used to assess the model. The performance metrics evaluates the performance after preprocessing, dataset splitting, feature selection, and classification. The evaluation of performance is measured individually for both datasets using the BEPO-KELM model. The model obtained 99.01% of accuracy, 98.38% of detection rate, 99.13% of specificity, 99.21% of precision, 99.05% of f-measure, and 0.068 of FPR using the CIC-IDS-2018 dataset. The model obtained 99.24% of accuracy, 99.17% of detection rate, 99.42% of specificity, 99.36% of precision, 99.18% of f-measure, and 0.055 of FPR using the Bot-IoT dataset.</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 systematic modelling of intrusion detection in IoT using a hybrid machine learning approach

  • 期刊论文

  • 发表日期：

    2025年9月9日

查看全文
  • M. Reji,
  • Christeena Joseph,
  • M. Sujatha,
  • M. Amanullah

摘要

Security issues have increased due to the increasing amount of IoT devices, making good intrusion detection systems (IDS) necessary to counteract changing cyberthreats. In order to overcome these obstacles, this work suggests combining a Kernel Extreme Learning Machine (KELM) method with a Binary Emperor Penguin Optimization (BEPO) algorithm-based feature selection for effective classification. Two benchmark datasets, CIC-IDS-2018 and Bot-IoT, were used to assess the model. The performance metrics evaluates the performance after preprocessing, dataset splitting, feature selection, and classification. The evaluation of performance is measured individually for both datasets using the BEPO-KELM model. The model obtained 99.01% of accuracy, 98.38% of detection rate, 99.13% of specificity, 99.21% of precision, 99.05% of f-measure, and 0.068 of FPR using the CIC-IDS-2018 dataset. The model obtained 99.24% of accuracy, 99.17% of detection rate, 99.42% of specificity, 99.36% of precision, 99.18% of f-measure, and 0.055 of FPR using the Bot-IoT dataset.