Scholarlink

<p>The problem of detecting intrusions in IoT-enabled sensor networks involves issues such as dealing with high dimensional data and continuously evolving threats. In this regard, this study suggests an algorithm known as Hybrid Optimized Boosted Bayesian Ensemble Model (OB²EM). It incorporates boosting-based ensemble learning technique along with Bayesian hyperparameter optimization and feature selection based on permutation importance. It makes use of the training of the base models in sequence and iteratively emphasizes misclassified samples. At the same time, it utilizes the Bayesian optimization technique to efficiently find an optimum combination of parameters. By doing so, the size of the dataset is reduced to 28 most important features. This model is tested on various benchmark datasets, including NSL-KDD, KDD CUP 99, and UNSW-NB15, which cover a variety of network traffic. The proposed OB²EM model shows impressive results with accuracy, precision, recall, and F1-score being equal to 98.36%, 98.25%, 98.22%, and 98.38% correspondingly. As compared to other classification techniques such as SVM, CNN, RNN, and GRU, the proposed algorithm works better.</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 OB²EM Framework with Bayesian Optimization for Real-Time Intrusion Detection in IoT Sensor Networks

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Chejarla Hari Kishore,
  • K. V. D. Kiran

摘要

The problem of detecting intrusions in IoT-enabled sensor networks involves issues such as dealing with high dimensional data and continuously evolving threats. In this regard, this study suggests an algorithm known as Hybrid Optimized Boosted Bayesian Ensemble Model (OB²EM). It incorporates boosting-based ensemble learning technique along with Bayesian hyperparameter optimization and feature selection based on permutation importance. It makes use of the training of the base models in sequence and iteratively emphasizes misclassified samples. At the same time, it utilizes the Bayesian optimization technique to efficiently find an optimum combination of parameters. By doing so, the size of the dataset is reduced to 28 most important features. This model is tested on various benchmark datasets, including NSL-KDD, KDD CUP 99, and UNSW-NB15, which cover a variety of network traffic. The proposed OB²EM model shows impressive results with accuracy, precision, recall, and F1-score being equal to 98.36%, 98.25%, 98.22%, and 98.38% correspondingly. As compared to other classification techniques such as SVM, CNN, RNN, and GRU, the proposed algorithm works better.