Scholarlink

<p>Industrial Internet of Things (IIoT) links different sensors, industrial programs, services, databases, machines, and individuals within the workforce. The IIoT improves our life by making cities smarter, agriculture, e-healthcare, etc. Detection of cyber-attacks in IIoT settings has special challenges because of the complexity of these networks, limitations of resources, and real-time nature of these networks. Conventional methods of detection are frequently not able to cope with the dynamism of IIoT. As an example, most of the current techniques are signature-based detection, which cannot detect the changing threats. Other methods like the anomaly based detection can create a high number of False Positives creating inefficiencies in managing threats. In order to address these concerns, a new Red Deer Optimized Dilated Convolutional Neural Network-based Bidirectional Gated Recurrent Unit (RDO-DCNet) was developed to help identify and stop malware attacks in the IIoT. The suggested method employs Red Deer Optimization (RDO) method to find optimal features and trim down dimensions of the dataset retaining important data. The relevant Dilated Convolutional Neural Network-based Bidirectional Gated Recurrent Unit (DCNN-BiGRU) architecture is applied to the accurate classification of Attack and Non-Attack categories, in addition to improving the classification efficiency with minimal computing complexity. A MATLAB simulator can be used to validate the performance of the RDO-DCNet using the real-time DS2OS dataset. The suggested RDO-DCNet framework has an accuracy of 99.31% compared to that of Fed-IIoTW method 83.32%, the AILBSM method 88.57% and the RAPTOR method 94.12% respectively. The optimized feature set also lowers computational complexity, enabling faster inference suitable for real-time industrial environments. Overall, the proposed framework offers a robust and efficient malware-detection solution for IIoT systems, demonstrating strong generalization, reduced overhead, and high reliability for protecting modern industrial operations.</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

RDO-DCNet: Red Deer Optimized Deep Learning Framework for Malware Detection in IIoT

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月22日

查看全文
  • Vijayaraj Alwarsamy,
  • Jebakumar Rethnaraj,
  • Mageshkumar Nagappan,
  • Lakshmi Patibandla Koteswara Rao Patibandla

摘要

Industrial Internet of Things (IIoT) links different sensors, industrial programs, services, databases, machines, and individuals within the workforce. The IIoT improves our life by making cities smarter, agriculture, e-healthcare, etc. Detection of cyber-attacks in IIoT settings has special challenges because of the complexity of these networks, limitations of resources, and real-time nature of these networks. Conventional methods of detection are frequently not able to cope with the dynamism of IIoT. As an example, most of the current techniques are signature-based detection, which cannot detect the changing threats. Other methods like the anomaly based detection can create a high number of False Positives creating inefficiencies in managing threats. In order to address these concerns, a new Red Deer Optimized Dilated Convolutional Neural Network-based Bidirectional Gated Recurrent Unit (RDO-DCNet) was developed to help identify and stop malware attacks in the IIoT. The suggested method employs Red Deer Optimization (RDO) method to find optimal features and trim down dimensions of the dataset retaining important data. The relevant Dilated Convolutional Neural Network-based Bidirectional Gated Recurrent Unit (DCNN-BiGRU) architecture is applied to the accurate classification of Attack and Non-Attack categories, in addition to improving the classification efficiency with minimal computing complexity. A MATLAB simulator can be used to validate the performance of the RDO-DCNet using the real-time DS2OS dataset. The suggested RDO-DCNet framework has an accuracy of 99.31% compared to that of Fed-IIoTW method 83.32%, the AILBSM method 88.57% and the RAPTOR method 94.12% respectively. The optimized feature set also lowers computational complexity, enabling faster inference suitable for real-time industrial environments. Overall, the proposed framework offers a robust and efficient malware-detection solution for IIoT systems, demonstrating strong generalization, reduced overhead, and high reliability for protecting modern industrial operations.