Scholarlink

<p>Industrial automation increasingly relies on interoperable Industrial Internet of Things (IIoT) devices, but ensuring secure communication across diverse protocols remains a challenge. Edge computing offers a solution by accelerating data analysis closer to devices. However, this introduces new vulnerabilities requiring vigilant network monitoring. To address this, we propose a comprehensive Network Intrusion Detection System (NIDS) framework tailored for edge-based IIoT environments. Our framework includes data preprocessing, sampling, feature selection, and a deep learning-based anomaly detection model. To alleviate the data sample crisis, we employed the Difficult Set Sampling for Training Efficiency (DSSTE) algorithm. A hybrid feature selection approach, leveraging the strengths of RF, XGBoost-based shuffling, filter, wrapper, and embedded methods, was used to identify critical features. Experimental results demonstrate the presented NIDS’s superior accuracy (99.9%) and low false alarm rate (0.04%) in detecting various attacks, even with a reduced feature set. The frameworks efficiency and effectiveness highlight its potential for enhancing the security of edge-based IIoT 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

Building robust lightweight network intrusion detection: a multi-pronged approach with DSSTE and deep learning for securing edge-enabled industrial IoT applications

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Mohemmed Yousuf Rahamathulla,
  • Mangayarkarasi Ramaiah

摘要

Industrial automation increasingly relies on interoperable Industrial Internet of Things (IIoT) devices, but ensuring secure communication across diverse protocols remains a challenge. Edge computing offers a solution by accelerating data analysis closer to devices. However, this introduces new vulnerabilities requiring vigilant network monitoring. To address this, we propose a comprehensive Network Intrusion Detection System (NIDS) framework tailored for edge-based IIoT environments. Our framework includes data preprocessing, sampling, feature selection, and a deep learning-based anomaly detection model. To alleviate the data sample crisis, we employed the Difficult Set Sampling for Training Efficiency (DSSTE) algorithm. A hybrid feature selection approach, leveraging the strengths of RF, XGBoost-based shuffling, filter, wrapper, and embedded methods, was used to identify critical features. Experimental results demonstrate the presented NIDS’s superior accuracy (99.9%) and low false alarm rate (0.04%) in detecting various attacks, even with a reduced feature set. The frameworks efficiency and effectiveness highlight its potential for enhancing the security of edge-based IIoT networks.