Scholarlink

<p>The emergence of 5G networks within the Internet of Things (IoT) networks is associated with both new opportunities and the increasing level of security risks. In this paper, it is hypothesized that a hybrid deep learning architecture between Convolutional Neural Networks (CNNs) and Bidirectional Gated Recurrent Units (Bi-GRUs) can be used to implement an intelligent intrusion detector. The proposed model based on the UNSW-NB15 dataset successfully captures both spatial and temporal patterns of network to determine malicious traffic and anomalies within 5G settings. CB-GRU hybrid model has shown better performance even when compared to LSTM and GRU architectures with 96.84% accuracy, 95.72% precision and a ROC-AUC score of 0.97, indicating increased performance of the model in terms of detection efficiency and robustness. The model automatically derives discriminative representations out of the complex network data unlike the traditional intrusion detection method that provides a set of handcrafted features used to identify attacks, the model is adaptive to counterattack trends. The research enables the development of secure infrastructures in IoT by enabling an intrusion detection system based on the data that is scalable to real-time protection of the 5G network.</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

Deep learning models for intrusion detection of irregularities and security breach detection in 5G networks

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Jyoti Srivastava,
  • Jay Prakash

摘要

The emergence of 5G networks within the Internet of Things (IoT) networks is associated with both new opportunities and the increasing level of security risks. In this paper, it is hypothesized that a hybrid deep learning architecture between Convolutional Neural Networks (CNNs) and Bidirectional Gated Recurrent Units (Bi-GRUs) can be used to implement an intelligent intrusion detector. The proposed model based on the UNSW-NB15 dataset successfully captures both spatial and temporal patterns of network to determine malicious traffic and anomalies within 5G settings. CB-GRU hybrid model has shown better performance even when compared to LSTM and GRU architectures with 96.84% accuracy, 95.72% precision and a ROC-AUC score of 0.97, indicating increased performance of the model in terms of detection efficiency and robustness. The model automatically derives discriminative representations out of the complex network data unlike the traditional intrusion detection method that provides a set of handcrafted features used to identify attacks, the model is adaptive to counterattack trends. The research enables the development of secure infrastructures in IoT by enabling an intrusion detection system based on the data that is scalable to real-time protection of the 5G network.