Scholarlink

<p>With the rapid deployment of 5G technology, industries are increasingly turning to private 5G networks to support mission-critical applications such as smart manufacturing, healthcare, and autonomous systems. While these networks offer enhanced control and performance, they are also more vulnerable to Distributed Denial of Service (DDoS) attacks, which can disrupt essential services by overwhelming network resources. This study aims to address this challenge by proposing a real-time DDoS detection and mitigation model tailored for private 5G environments. Specifically, we develop a hybrid deep learning model that combines Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (Bi-LSTM) networks. The CNN extracts spatial patterns in traffic, while the Bi-LSTM captures temporal dependencies, enabling accurate classification of network behavior. We simulate a 5G private network using free5GC and UERANSIM, and generate both benign and malicious traffic using iperf3 and hping3. The model is trained and evaluated on the CICDDoS2019 dataset, which contains 13 types of real-world DDoS attack scenarios. Key results show that our model achieves: Accuracy: 99.72%, Precision: 99.51%, 99.92% and F1-score: 99.71%. Additionally, the model demonstrates strong real-time mitigation capabilities through Linux Traffic Control (TC), reducing attack traffic by over 80% upon detection. These results validate the effectiveness of our CNN–BiLSTM model in enhancing 5G network resilience, and highlight its practical applicability in securing modern, latency-sensitive private networks.</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

Mitigating DDoS Attacks in Private 5G Networks Using Deep Learning

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • Chien-Liang Chen,
  • Yin-Jing Rao,
  • Chun-Hsien Sung

摘要

With the rapid deployment of 5G technology, industries are increasingly turning to private 5G networks to support mission-critical applications such as smart manufacturing, healthcare, and autonomous systems. While these networks offer enhanced control and performance, they are also more vulnerable to Distributed Denial of Service (DDoS) attacks, which can disrupt essential services by overwhelming network resources. This study aims to address this challenge by proposing a real-time DDoS detection and mitigation model tailored for private 5G environments. Specifically, we develop a hybrid deep learning model that combines Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (Bi-LSTM) networks. The CNN extracts spatial patterns in traffic, while the Bi-LSTM captures temporal dependencies, enabling accurate classification of network behavior. We simulate a 5G private network using free5GC and UERANSIM, and generate both benign and malicious traffic using iperf3 and hping3. The model is trained and evaluated on the CICDDoS2019 dataset, which contains 13 types of real-world DDoS attack scenarios. Key results show that our model achieves: Accuracy: 99.72%, Precision: 99.51%, 99.92% and F1-score: 99.71%. Additionally, the model demonstrates strong real-time mitigation capabilities through Linux Traffic Control (TC), reducing attack traffic by over 80% upon detection. These results validate the effectiveness of our CNN–BiLSTM model in enhancing 5G network resilience, and highlight its practical applicability in securing modern, latency-sensitive private networks.