A multi-stream deep learning architecture for DDoS detection in IoT networks

Abstract The proliferation of Internet of Things (IoT) devices has led to a surge in network vulnerabilities, particularly to Distributed Denial-of-Service (DDoS) attacks, which can severely disrupt critical services. To address this challenge, we propose HybNet-HD, a novel multi-stream deep learning architecture tailored for effective DDoS detection in IoT environments. The framework combines convolutional feature extraction for capturing spatial hierarchies, dense relational learning to model global dependencies, and a fusion strategy guided by Hellinger Distance-based dissimilarity and attention mechanisms. This design enables the network to focus on the most informative features while maintaining robustness across diverse attack patterns. Regularization techniques such as dropout, early stopping, and adaptive learning rates are incorporated to enhance generalization. Comprehensive evaluations on benchmark datasets–UNSW-NB15, NSL-KDD, and LATAM-IoT DDoS–demonstrate that HybNet-HD achieves superior performance compared to existing methods in terms of evaluation measures, and inference efficiency, making it a reliable solution for real-time DDoS detection in heterogeneous IoT ecosystems.