Scholarlink

This paper presents an innovative neural network architecture designed to optimize network traffic data processing. The architecture integrates two novel approaches: a Fourier Transform-based method to adjust network layer depth and an attention mechanism driven by the information gain ratio of API calls. Experimental results on datasets like NSL-KDD and CICIDS2017 show that our model outperforms traditional methods, including ARIMA, in F1-score and training efficiency. Ablation studies confirm the effectiveness of both innovations, highlighting their contribution to improved model performance.

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

A Deep Learning-Based Method for Network Anomaly Traffic Detection Using Fourier Transform and Information Gain-Optimized Attention Mechanism

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ruitong Liu,
  • Yiyang Xiong,
  • Haoming Zhang,
  • Fangning Shi,
  • Junsheng Mu

摘要

This paper presents an innovative neural network architecture designed to optimize network traffic data processing. The architecture integrates two novel approaches: a Fourier Transform-based method to adjust network layer depth and an attention mechanism driven by the information gain ratio of API calls. Experimental results on datasets like NSL-KDD and CICIDS2017 show that our model outperforms traditional methods, including ARIMA, in F1-score and training efficiency. Ablation studies confirm the effectiveness of both innovations, highlighting their contribution to improved model performance.