Scholarlink

<p>Advancements in wireless technology have exposed the Industrial Internet of Things (IIoT) ecosystem to cybersecurity risks. These vulnerabilities provide significant threats, including harm to production systems, and substantial monetary losses. This research introduces a deep hybrid learning model to enhance cyberattack detection and protect resilient data inside an asynchronous federated learning environment. Integrating attention-enhanced Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) architectures offer an excellent method for efficiently detecting abnormalities in data from IIoT sensors. Our approach functions asynchronously to improve security, keeping content localized, and avoiding the need for complete network synchronization. The model’s effectiveness is evaluated using two well-known datasets: Edge-IIoTset and CICIoT 2023. The results indicate exceptional performance, with the model attaining a flawless 1.00 in accuracy, precision, recall, and F1-score on both datasets. The trial findings underscore the model’s remarkable flexibility and capacity to swiftly address emergent risks, representing a substantial advancement in safeguarding IIoT infrastructures and ensuring stringent data privacy preservation.</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

Protecting IIoT ecosystems: a hybrid CNN-LSTM approach in federated learning

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Qawsar Gulzar,
  • Khurram Mustafa

摘要

Advancements in wireless technology have exposed the Industrial Internet of Things (IIoT) ecosystem to cybersecurity risks. These vulnerabilities provide significant threats, including harm to production systems, and substantial monetary losses. This research introduces a deep hybrid learning model to enhance cyberattack detection and protect resilient data inside an asynchronous federated learning environment. Integrating attention-enhanced Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) architectures offer an excellent method for efficiently detecting abnormalities in data from IIoT sensors. Our approach functions asynchronously to improve security, keeping content localized, and avoiding the need for complete network synchronization. The model’s effectiveness is evaluated using two well-known datasets: Edge-IIoTset and CICIoT 2023. The results indicate exceptional performance, with the model attaining a flawless 1.00 in accuracy, precision, recall, and F1-score on both datasets. The trial findings underscore the model’s remarkable flexibility and capacity to swiftly address emergent risks, representing a substantial advancement in safeguarding IIoT infrastructures and ensuring stringent data privacy preservation.