Scholarlink

<p>The rapid proliferation of internet of things (IoT) devices introduces considerable security challenges, including device heterogeneity, large-scale traffic, and constantly evolving attack surfaces. Conventional security mechanisms often fail to address these complexities, whereas intrusion detection systems (IDSs) play a crucial role in providing intelligent monitoring and adaptive defense against diverse threats. To this end, this study presents an advanced IDS that integrates SHAP-based feature selection with ensemble learning to enhance detection performance. Leveraging the CICIDS2017 and CICIoT2023 benchmark datasets, the proposed two-stage framework initially employs feature selection to extract the most discriminative attributes, followed by an ensemble of classifiers to ensure robust prediction. Experimental analysis demonstrates its superior effectiveness, attaining 99.99% and 97.58% accuracies in multiclass attack detection, while markedly reducing false alarm rates and outperforming contemporary state-of-the-art techniques.</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

Securing IoT Environment Against Intrusions Using Voting Ensemble Learning Classifier

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Aishwarya Vardhan,
  • Prashant Kumar,
  • Lalit Kumar Awasthi

摘要

The rapid proliferation of internet of things (IoT) devices introduces considerable security challenges, including device heterogeneity, large-scale traffic, and constantly evolving attack surfaces. Conventional security mechanisms often fail to address these complexities, whereas intrusion detection systems (IDSs) play a crucial role in providing intelligent monitoring and adaptive defense against diverse threats. To this end, this study presents an advanced IDS that integrates SHAP-based feature selection with ensemble learning to enhance detection performance. Leveraging the CICIDS2017 and CICIoT2023 benchmark datasets, the proposed two-stage framework initially employs feature selection to extract the most discriminative attributes, followed by an ensemble of classifiers to ensure robust prediction. Experimental analysis demonstrates its superior effectiveness, attaining 99.99% and 97.58% accuracies in multiclass attack detection, while markedly reducing false alarm rates and outperforming contemporary state-of-the-art techniques.