Scholarlink

<p>Cloud-based music education platforms are increasingly exposed to sophisticated cyberattacks, requiring intrusion detection systems capable of handling high-dimensional and dynamic traffic. This paper proposes a novel Residual Knowledge Graph Meerkat Group Attention Network (Res-K2G-MAN) for effective intrusion detection in such environments. The framework integrates Knowledge Graph-based Graph Neural Networks (KG-GNN) with a Residual Group Attention Network (ResGANet) to capture structural and semantic dependencies in traffic flows. An Uncertainty-Aware Decision Transformer (UADT) is employed for feature extraction, while the Meerkat Optimization Algorithm (MOA) enhances parameter tuning and convergence. Experiments conducted on the Cloud_CICIDS2017 and NSL-KDD_MusicCloud datasets demonstrate that the proposed model achieves superior accuracy and robustness compared to existing methods, particularly in detecting low-frequency and stealthy intrusions. Results from experiments show that the suggested strategy accomplishes an outstanding 99.9% accuracy in detecting intrusions. The results establish Res-K2G-MAN as a scalable and adaptive solution for securing online education systems hosted on public cloud infrastructures. Key advantages of this method include enhanced detection of low-frequency attacks and improved adaptability to dynamic cloud-based environments.</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

Res-K2G-MAN: a residual knowledge graph attention framework for intrusion detection in online music education over public cloud networks

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • V. Ravikumar,
  • G. Umarani Srikanth,
  • Manigandan S K,
  • A. Vishnukumar

摘要

Cloud-based music education platforms are increasingly exposed to sophisticated cyberattacks, requiring intrusion detection systems capable of handling high-dimensional and dynamic traffic. This paper proposes a novel Residual Knowledge Graph Meerkat Group Attention Network (Res-K2G-MAN) for effective intrusion detection in such environments. The framework integrates Knowledge Graph-based Graph Neural Networks (KG-GNN) with a Residual Group Attention Network (ResGANet) to capture structural and semantic dependencies in traffic flows. An Uncertainty-Aware Decision Transformer (UADT) is employed for feature extraction, while the Meerkat Optimization Algorithm (MOA) enhances parameter tuning and convergence. Experiments conducted on the Cloud_CICIDS2017 and NSL-KDD_MusicCloud datasets demonstrate that the proposed model achieves superior accuracy and robustness compared to existing methods, particularly in detecting low-frequency and stealthy intrusions. Results from experiments show that the suggested strategy accomplishes an outstanding 99.9% accuracy in detecting intrusions. The results establish Res-K2G-MAN as a scalable and adaptive solution for securing online education systems hosted on public cloud infrastructures. Key advantages of this method include enhanced detection of low-frequency attacks and improved adaptability to dynamic cloud-based environments.