Scholarlink

As the Industrial Internet of Things (IIoT) expands in scope and application, Cyber-Physical Systems (CPS) in industrial environments are increasingly vulnerable to cyber threats. This paper presents an integrated anomaly detection model by combining autoencoders with decision tree-based classification (AutoDTClass) to detect and mitigate cyberattacks in IIoT networks, a critical subset of CPS, through a data-driven approach using machine learning. The research utilises the Edge-IIoT dataset, which was selected for its comprehensive representation of industrial network attacks. The proposed model achieves high accuracy in anomaly detection, allowing the system to differentiate between cyber threats while adapting to new data patterns. Therefore it can improve anomaly detection accuracy in real-time IIoT environments, contributing to a resilient security layer for CPS amidst increasing operational complexity.

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

AutoDTClass: Real-Time Anomaly Detection and Threat Classification in IIoT Networks Using Autoencoders and Decision Trees

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jakub Możaryn,
  • Cezary Krzemiński,
  • Cédric Escudero

摘要

As the Industrial Internet of Things (IIoT) expands in scope and application, Cyber-Physical Systems (CPS) in industrial environments are increasingly vulnerable to cyber threats. This paper presents an integrated anomaly detection model by combining autoencoders with decision tree-based classification (AutoDTClass) to detect and mitigate cyberattacks in IIoT networks, a critical subset of CPS, through a data-driven approach using machine learning. The research utilises the Edge-IIoT dataset, which was selected for its comprehensive representation of industrial network attacks. The proposed model achieves high accuracy in anomaly detection, allowing the system to differentiate between cyber threats while adapting to new data patterns. Therefore it can improve anomaly detection accuracy in real-time IIoT environments, contributing to a resilient security layer for CPS amidst increasing operational complexity.