Scholarlink

<p>Cyber-physical systems (CPS) in water treatment facilities are increasingly vulnerable to Internet of Things (IoT)-based attacks, with statistics showing a 47% rise in malicious activities targeting critical infrastructure and 62% of such attacks leading to significant operational disruptions. Existing methods to detect and classify these attacks often face challenges related to high false-positive rates and difficulty distinguishing complex attack patterns. To address these issues, this study proposes a Novel Cyber-Physical Attack Classification (CPAC-Net) framework, which begins with comprehensive data preprocessing, including normalization and outlier detection, to ensure clean input for analysis. Feature extraction is performed using Principal Decision Component Tree Analysis (PDCTA), which reduces dimensionality while retaining critical attributes from the Secure Water Treatment (SWaT) dataset’s sensor and actuator readings. These features are then fed into a Stacked Autoencoder Adopted Deep Neural Network (SAA-DNN) for classification, designed to differentiate between seven attack classes. The proposed framework performs better in accurately identifying simple and sophisticated cyber-physical threats in IoT-enabled water treatment systems.</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

CPAC-Net: deep learning approach for cyber-physical attack classification in secure water treatment data

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月1日

查看全文
  • Chelli Nithish,
  • T. Rajesh,
  • D. Bhagyalaxmi,
  • Rendla Aravind,
  • Gillala Anshuman,
  • P. Rithin Yadav

摘要

Cyber-physical systems (CPS) in water treatment facilities are increasingly vulnerable to Internet of Things (IoT)-based attacks, with statistics showing a 47% rise in malicious activities targeting critical infrastructure and 62% of such attacks leading to significant operational disruptions. Existing methods to detect and classify these attacks often face challenges related to high false-positive rates and difficulty distinguishing complex attack patterns. To address these issues, this study proposes a Novel Cyber-Physical Attack Classification (CPAC-Net) framework, which begins with comprehensive data preprocessing, including normalization and outlier detection, to ensure clean input for analysis. Feature extraction is performed using Principal Decision Component Tree Analysis (PDCTA), which reduces dimensionality while retaining critical attributes from the Secure Water Treatment (SWaT) dataset’s sensor and actuator readings. These features are then fed into a Stacked Autoencoder Adopted Deep Neural Network (SAA-DNN) for classification, designed to differentiate between seven attack classes. The proposed framework performs better in accurately identifying simple and sophisticated cyber-physical threats in IoT-enabled water treatment systems.