Scholarlink

<p>Although the Software-Defined Internet of Drones (SD-IoD) integrates the Internet of Drones with Software-Defined Network for flexible, dynamic drone communications, it remains vulnerable to Distributed Denial of Service (DDoS) attacks that can exhaust resources and disrupt operations. Therefore, the paper proposes a novel Semi-Supervised Federated Learning with Hyperparameter Tuning (SsFLHT) framework for efficiently and accurately detecting DDoS attacks in SD-IoD environments. SsFLHT leverages semi-supervised learning, which iteratively assigns pseudo-labels to unlabelled live network traffic based on a confidence threshold. Moreover, the method introduces an early phase of federated hyperparameter tuning (HT) using the modified Greedy Sand Cat Optimisation (GSCO) and an adaptive early stopping mechanism, which together accelerate convergence during both local and global training phases. GSCO mimics a sand-cat by sniffing multiple promising prey scents, then repeatedly narrowing its observation, tuning its hearing bias, and pouncing on the clearest prey location. Experimental results show that SsFLHT improves convergence speed and detection accuracy, outperforming traditional centralised, federated learning baselines, and the other HT methods. The method achieves global detection accuracy of 96.21% with a Multi-Layer Perceptron (MLP) and 95.27% with a Convolutional Neural Network (CNN), demonstrating its effectiveness in securing SD-IoD networks against DDoS attacks.</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

DDoS attack detection using semi-supervised hyperparameter optimized federated learning in software defined internet of drones

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Fauzi Dwi Setiawan Sumadi,
  • Khalid Alsubhi

摘要

Although the Software-Defined Internet of Drones (SD-IoD) integrates the Internet of Drones with Software-Defined Network for flexible, dynamic drone communications, it remains vulnerable to Distributed Denial of Service (DDoS) attacks that can exhaust resources and disrupt operations. Therefore, the paper proposes a novel Semi-Supervised Federated Learning with Hyperparameter Tuning (SsFLHT) framework for efficiently and accurately detecting DDoS attacks in SD-IoD environments. SsFLHT leverages semi-supervised learning, which iteratively assigns pseudo-labels to unlabelled live network traffic based on a confidence threshold. Moreover, the method introduces an early phase of federated hyperparameter tuning (HT) using the modified Greedy Sand Cat Optimisation (GSCO) and an adaptive early stopping mechanism, which together accelerate convergence during both local and global training phases. GSCO mimics a sand-cat by sniffing multiple promising prey scents, then repeatedly narrowing its observation, tuning its hearing bias, and pouncing on the clearest prey location. Experimental results show that SsFLHT improves convergence speed and detection accuracy, outperforming traditional centralised, federated learning baselines, and the other HT methods. The method achieves global detection accuracy of 96.21% with a Multi-Layer Perceptron (MLP) and 95.27% with a Convolutional Neural Network (CNN), demonstrating its effectiveness in securing SD-IoD networks against DDoS attacks.