A Scalable Hybrid CNN-LSTM and SDN Based Architecture for Real-Time DDoS Defence in IoT and Cloud Networks

Abstract

Distributed Denial-of-Service (DDoS) is one of the most significant threats to the security of IoT, cloud, and Software-Defined Network (SDN) environments because it can aggravate network unavailability and disrupt the continuity of services. This study presents a hybrid CNN-LSTM-based DDoS detection and mitigation system for real-time DDoS detection in the context of heterogeneous network environments, which is integrated with SDN and cloud-edge collaborative architecture. The proposed framework is based on two deep learning models, Convolutional Neural Networks (CNN) for extracting spatial traffic features and Long Short-Term Memory (LSTM) networks for learning temporal dependencies, to achieve better detection of complex attack patterns that involve multiple vectors. To evaluate the experimental results, the CIC-DDoS2019 dataset, along with some other benchmark traffic datasets, is used under heterogeneous deployment conditions. The proposed Hybrid CNN-LSTM framework was able to achieve an accuracy of 95.88% and an F1-score of 91.41%, while the lightweight CNN models for the IoT environments had an accuracy of more than 94% with lower latency and resource usage. The results show that the proposed framework effectively detects DDoS attacks in real time, highly adapts to network changes, and has high scalability of deployment for SDN, cloud, and IoT environments.