Scholarlink

This paper introduces an anomaly detection system using the Activity and Event Network (AEN) model, Graph Neural Networks (GNN), and Graph Attention Networks (GAT). Our method gives more attention to neighbors with higher in-degree, highlighting their importance in the network. By combining the dynamic nature of AEN with the attention mechanisms of GAT, we can better identify anomalies in network behavior over time. Our approach effectiveness was assessed using two datasets and resulted in an accuracy of 88.3% and 90.7%.

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 Anomaly Detection System Using an Attention-Based GNN

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amir Mohammadi Bagha,
  • Isaac Woungang,
  • Issa Traore,
  • Danda B. Rawat,
  • Sudeep Tanwar,
  • Azin Hassanalizadeh

摘要

This paper introduces an anomaly detection system using the Activity and Event Network (AEN) model, Graph Neural Networks (GNN), and Graph Attention Networks (GAT). Our method gives more attention to neighbors with higher in-degree, highlighting their importance in the network. By combining the dynamic nature of AEN with the attention mechanisms of GAT, we can better identify anomalies in network behavior over time. Our approach effectiveness was assessed using two datasets and resulted in an accuracy of 88.3% and 90.7%.