Scholarlink

<p>The rapid spread of the Industrial Internet of Things (IIoT) has greatly improved automation and operational efficiency. However, the increased connectivity of IIoT devices imposes serious cybersecurity challenges, and robust anomaly detection methods are required to address these. This paper develops an optimized Light Gradient Boosting Machine model for the prediction of anomalies in IIoT networks, using the communication patterns and network traffic behaviours. The proposed model integrates with hyperparameter optimization using an evolutionary algorithm (Self-adaptive Differential Evolution (SADE)) to improve the accuracy of detection and reduce computational complexity using X-IIoTID dataset. Comparison of the proposed anomaly detection model with traditional machine learning model based on recall,precision,F-beta and F1-score. Further, the adopted approach of hyperparameter optimization has been compared with variants of DE and other benchmark optimization techniques. The proposed model indicates good high detection rate as well as less false positives are made for an appropriate application on real-time anomaly detection within an IIoT environment.</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

Communication and Traffic Patterns-Aware Evolutionary Optimized Light Gradient Boosting Model for Anomaly Prediction in Industrial IoT

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Geetanjali Bhoi,
  • Bighnaraj Naik,
  • Rajat Kumar Sahu,
  • Etuari Oram

摘要

The rapid spread of the Industrial Internet of Things (IIoT) has greatly improved automation and operational efficiency. However, the increased connectivity of IIoT devices imposes serious cybersecurity challenges, and robust anomaly detection methods are required to address these. This paper develops an optimized Light Gradient Boosting Machine model for the prediction of anomalies in IIoT networks, using the communication patterns and network traffic behaviours. The proposed model integrates with hyperparameter optimization using an evolutionary algorithm (Self-adaptive Differential Evolution (SADE)) to improve the accuracy of detection and reduce computational complexity using X-IIoTID dataset. Comparison of the proposed anomaly detection model with traditional machine learning model based on recall,precision,F-beta and F1-score. Further, the adopted approach of hyperparameter optimization has been compared with variants of DE and other benchmark optimization techniques. The proposed model indicates good high detection rate as well as less false positives are made for an appropriate application on real-time anomaly detection within an IIoT environment.