Scholarlink

<p>Internet of Things intrusion detection involves monitoring devices and network traffic to identify security threats or malicious activities. Graph Neural Network-based methods address the limitations of traditional NIDS in capturing spatial topology information but face challenges: (1) ineffective initialization of node features during graph construction, (2) underutilization of edge features. To overcome these issues, we propose an edge-enhanced GraphSAGE-based IoT intrusion detection model. First, we assign features to nodes by aggregating the features of their adjacent edges, generating expressive node features. Then, we employ a doubling operation to integrate both node and edge features during message aggregation. Finally, inspired by residual connections, we preserve original edge attributes during edge feature updates to prevent over-smoothing. Experimental results demonstrate that the proposed model achieves F1 scores of 99.41%, 98.66%, and 99.78% on three public datasets: CICIoT2023, Edge-IIoT, and BoT-IoT, These results indicate the model’s superior classification performance in IoT environments and outperforming existing methods. The code is available at <a href="https://github.com/shendazhuang/EE-GraphSAGE">https://github.com/shendazhuang/EE-GraphSAGE</a>.</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

An edge-enhanced graphSAGE-based intrusion detection model for the internet of things

  • 期刊论文

  • 发表日期：

    2025年4月28日

查看全文
  • Ruisheng Li,
  • Huimin Shen,
  • Qilong Zhang,
  • Huimin Duan

摘要

Internet of Things intrusion detection involves monitoring devices and network traffic to identify security threats or malicious activities. Graph Neural Network-based methods address the limitations of traditional NIDS in capturing spatial topology information but face challenges: (1) ineffective initialization of node features during graph construction, (2) underutilization of edge features. To overcome these issues, we propose an edge-enhanced GraphSAGE-based IoT intrusion detection model. First, we assign features to nodes by aggregating the features of their adjacent edges, generating expressive node features. Then, we employ a doubling operation to integrate both node and edge features during message aggregation. Finally, inspired by residual connections, we preserve original edge attributes during edge feature updates to prevent over-smoothing. Experimental results demonstrate that the proposed model achieves F1 scores of 99.41%, 98.66%, and 99.78% on three public datasets: CICIoT2023, Edge-IIoT, and BoT-IoT, These results indicate the model’s superior classification performance in IoT environments and outperforming existing methods. The code is available at https://github.com/shendazhuang/EE-GraphSAGE.