Scholarlink

<p>In recent years, integrating Industrial Control Systems (ICSs) with internet-connected environments has heightened their exposure to cyberattacks. Despite the enhanced cybersecurity measures, emerging attacks targeting network traffic in Supervisory Control and Data Acquisition (SCADA) systems pose critical risks to the operational reliability of an ICS. Addressing these concerns requires an intelligent and adaptive defensive strategy, where Deep Learning (DL) approaches present a viable solution for identifying such complex attacks. Hence, this paper proposes a DL-based attack detection approach, HGCNN-LSTM to defend the security of ICS networks against data injection, Denial of Service attacks, and reconnaissance attacks. The proposed model uses hypergraphs to optimize the thresholds of the Convolutional Neural Network (CNN) based Long Short Term Memory (LSTM) architecture enhancing the attack detection performance. Extensive experiments on the publicly available Secure Water Treatment (SWaT) dataset demonstrate that the proposed attack detection framework achieves a better detection rate with minimal computational overhead. The proposed model successfully detected 31 out of 36 attacks outperforming the existing attack detection approaches with a recall value of 96.89% and an accuracy of 97.41%. The results underscore that DL-based approaches enhance the resilience of ICS networks against sophisticated cyber-attacks.</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

HGCNN-LSTM: A Data-driven Approach for Cyberattack Detection in Cyber-Physical Systems

  • 期刊论文

  • 发表日期：

    2025年1月7日

查看全文
  • S. Abinash,
  • N. Srivatsan,
  • S. K. Hemachandran,
  • S. Priyanga

摘要

In recent years, integrating Industrial Control Systems (ICSs) with internet-connected environments has heightened their exposure to cyberattacks. Despite the enhanced cybersecurity measures, emerging attacks targeting network traffic in Supervisory Control and Data Acquisition (SCADA) systems pose critical risks to the operational reliability of an ICS. Addressing these concerns requires an intelligent and adaptive defensive strategy, where Deep Learning (DL) approaches present a viable solution for identifying such complex attacks. Hence, this paper proposes a DL-based attack detection approach, HGCNN-LSTM to defend the security of ICS networks against data injection, Denial of Service attacks, and reconnaissance attacks. The proposed model uses hypergraphs to optimize the thresholds of the Convolutional Neural Network (CNN) based Long Short Term Memory (LSTM) architecture enhancing the attack detection performance. Extensive experiments on the publicly available Secure Water Treatment (SWaT) dataset demonstrate that the proposed attack detection framework achieves a better detection rate with minimal computational overhead. The proposed model successfully detected 31 out of 36 attacks outperforming the existing attack detection approaches with a recall value of 96.89% and an accuracy of 97.41%. The results underscore that DL-based approaches enhance the resilience of ICS networks against sophisticated cyber-attacks.