Scholarlink

<p>Industrial cyber–physical systems (CPSs) represent the convergence of traditional industrial control systems with advanced technologies like IoT (Internet of Things), cloud computing, and AI (artificial intelligence), enabling smart services, big data analytics, and efficient resource management. Current intrusion detection methods for industrial CPSs face limitations due to the scarcity of high-quality attack data and privacy concerns. To address these challenges, this paper proposes a hybrid deep learning framework that combines CNN-GRU-LSTM (convolutional neural network-gated recurrent unit-long short term memory) based intrusion detection framework for classification of cyber threats. CNN is used for spatial extraction of features from the pre-processed dataset followed by temporal features extraction using GRU. Finally, LSTM receives the vectored characteristics of the extracted features for classification of several attacks or anomalies. The performance of the proposed framework is evaluated and compared with existing studies based on accuracy (%), sensitivity (%) and specificity (%).</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

Hybrid deep learning framework for detection of anomalies in cyber-physical systems

  • 期刊论文

  • 发表日期：

    2025年7月18日

查看全文
  • Mohamed M. Abbassy,
  • A. Satya Sai Kumar,
  • Waleed M. Ead,
  • Ayman Aboalndr Mohamed Aboalndr,
  • Fahad Kamal Alsheref,
  • Mohammed Abdalla,
  • Aakanksha Shivani

摘要

Industrial cyber–physical systems (CPSs) represent the convergence of traditional industrial control systems with advanced technologies like IoT (Internet of Things), cloud computing, and AI (artificial intelligence), enabling smart services, big data analytics, and efficient resource management. Current intrusion detection methods for industrial CPSs face limitations due to the scarcity of high-quality attack data and privacy concerns. To address these challenges, this paper proposes a hybrid deep learning framework that combines CNN-GRU-LSTM (convolutional neural network-gated recurrent unit-long short term memory) based intrusion detection framework for classification of cyber threats. CNN is used for spatial extraction of features from the pre-processed dataset followed by temporal features extraction using GRU. Finally, LSTM receives the vectored characteristics of the extracted features for classification of several attacks or anomalies. The performance of the proposed framework is evaluated and compared with existing studies based on accuracy (%), sensitivity (%) and specificity (%).