Scholarlink

<p>The Industrial Internet of Things (IIoT) combines sensors, machinery, industrial software, databases, services, and personnel within the computing environment. IIoT enhances cities, agriculture, e-healthcare, and other areas of life. While IIoT and IoT share many characteristics, they deploy different cybersecurity methods for their networks. Unlike consumer IoT solutions, Industrial Internet of Things (IIoT) solutions typically integrate into operational systems instead of operating independently. Previous investigations suffered a major weakness when they failed to use sampling techniques to balance the datasets from real IIoT operations, which comprise multiple attack types. This can result in models performing poorly on minority classes but well on majority classes, limiting the intrusion detection system’s overall effectiveness. Therefore, security solutions for IIoT require extra strategic planning and continuous monitoring to protect system security and privacy. The objective is to identify traffic data irregularities from IIoT networks. This research introduces a deepCLG hybrid learning model designed to improve network intrusion detection systems (NIDSs). The datasets first go through preprocessing and normalization. Following this, we then formulated a hybrid learning model named DeepCLG. It integrates the convolutional neural network (CNN), the long short-term memory (LSTM), and the gated recurrent unit (GRU) with the capsule network (CN). The model’s efficacy is assessed using two widely accessible datasets: the CICIoT 2023 and UNSW_NB15 datasets. The proposed model outperformed existing techniques in detecting attacks and achieved an accuracy of 99.82%, precision of 97.83%, detection rate of 95.91%, Matthew’s correlation coefficient (MCC) of 96.77%, f1-score of 96.86%, and false alarm rate of 00.06% for the CICIoT 2023 dataset and an accuracy of 95.55%, precision of 88.78%, detection rate of 57.77%, Matthew’s correlation coefficient (MCC) of 79.26%, f1-score of 70.06%, and false alarm rate of 00.81% for the UNSW_NB15 dataset. These findings reveal that the proposed model is proficient in identifying cyberattacks and exhibits adaptability in detecting a multitude of cyberattacks within real IIoT environments.</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

Enhancing network security in industrial IoT environments: a DeepCLG hybrid learning model for cyberattack detection

  • 期刊论文

  • 发表日期：

    2025年1月24日

查看全文
  • Qawsar Gulzar,
  • Khuram Mustafa

摘要

The Industrial Internet of Things (IIoT) combines sensors, machinery, industrial software, databases, services, and personnel within the computing environment. IIoT enhances cities, agriculture, e-healthcare, and other areas of life. While IIoT and IoT share many characteristics, they deploy different cybersecurity methods for their networks. Unlike consumer IoT solutions, Industrial Internet of Things (IIoT) solutions typically integrate into operational systems instead of operating independently. Previous investigations suffered a major weakness when they failed to use sampling techniques to balance the datasets from real IIoT operations, which comprise multiple attack types. This can result in models performing poorly on minority classes but well on majority classes, limiting the intrusion detection system’s overall effectiveness. Therefore, security solutions for IIoT require extra strategic planning and continuous monitoring to protect system security and privacy. The objective is to identify traffic data irregularities from IIoT networks. This research introduces a deepCLG hybrid learning model designed to improve network intrusion detection systems (NIDSs). The datasets first go through preprocessing and normalization. Following this, we then formulated a hybrid learning model named DeepCLG. It integrates the convolutional neural network (CNN), the long short-term memory (LSTM), and the gated recurrent unit (GRU) with the capsule network (CN). The model’s efficacy is assessed using two widely accessible datasets: the CICIoT 2023 and UNSW_NB15 datasets. The proposed model outperformed existing techniques in detecting attacks and achieved an accuracy of 99.82%, precision of 97.83%, detection rate of 95.91%, Matthew’s correlation coefficient (MCC) of 96.77%, f1-score of 96.86%, and false alarm rate of 00.06% for the CICIoT 2023 dataset and an accuracy of 95.55%, precision of 88.78%, detection rate of 57.77%, Matthew’s correlation coefficient (MCC) of 79.26%, f1-score of 70.06%, and false alarm rate of 00.81% for the UNSW_NB15 dataset. These findings reveal that the proposed model is proficient in identifying cyberattacks and exhibits adaptability in detecting a multitude of cyberattacks within real IIoT environments.