Scholarlink

<p>To enhance cybersecurity in petroleum industrial control systems, this study proposes a network traffic anomaly detection model based on a Deep Belief Network (DBN) optimized by an Improved Grey Wolf Optimization (IGWO) algorithm. Network traffic data are collected using a Sniffer Data Flow Capture Tool, and four feature attributes - source and destination IP addresses along with source and destination port numbers - are extracted via information entropy to serve as inputs to the DBN. The IGWO algorithm optimizes the weights and biases of the DBN, thereby improving detection accuracy. Experimental results validate the effectiveness of the proposed model: under normal conditions, the feature entropy ranges from 0.55 to 0.6, whereas during attacks, the entropy decreases notably in the destination IP and port attributes. The model processes each data link within 0.36&#xa0;s, with outputs closely aligning with actual results. Across various attack scenarios, it achieves high accuracy (0.99), precision (0.97), and recall (0.98), significantly strengthening network security through rapid threat identification.</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

Network traffic anomaly detection model of petroleum industrial control software based on improved grey wolf optimization algorithm

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月1日

查看全文
  • Bo Yang,
  • Yang Xiao,
  • Xinzhang Wang,
  • Cheng Peng,
  • Xuqing Liu

摘要

To enhance cybersecurity in petroleum industrial control systems, this study proposes a network traffic anomaly detection model based on a Deep Belief Network (DBN) optimized by an Improved Grey Wolf Optimization (IGWO) algorithm. Network traffic data are collected using a Sniffer Data Flow Capture Tool, and four feature attributes - source and destination IP addresses along with source and destination port numbers - are extracted via information entropy to serve as inputs to the DBN. The IGWO algorithm optimizes the weights and biases of the DBN, thereby improving detection accuracy. Experimental results validate the effectiveness of the proposed model: under normal conditions, the feature entropy ranges from 0.55 to 0.6, whereas during attacks, the entropy decreases notably in the destination IP and port attributes. The model processes each data link within 0.36 s, with outputs closely aligning with actual results. Across various attack scenarios, it achieves high accuracy (0.99), precision (0.97), and recall (0.98), significantly strengthening network security through rapid threat identification.