IoT Intrusion Detection Using SDN Cloud-Edge Collaboration and Ensemble Deep Learning
Yuhua Xu, Yan Li, Zhixin Sun · 2024
5G technology promotes the development of Internet of Things (IoT), but IoT faces various attacks, which lead to cybersecurity threats and property damage. Software Defined Networking (SDN) is one of the key technologies of 5G, and it is a new network paradigm that facilitates monitoring network flows, enabling programmable networks. Leveraging SDN to implement intrusion detection can improve the security of local IoT. In order to ensure the reliable communication of IoT and realize the precise network protection, this paper proposes an IoT intrusion detection scheme based on SDN cloud-edge collaboration and Ensemble Deep Learning (EDL). First, the edge controller realizes the accurate classification of normal traffic and suspicious traffic through the joint model of Naive Bayesian and Back Propagation neural networks (NB - BP), and then transmits the suspicious traffic to the cloud for detection. The attack type of suspicious traffic is identified by the EDL model deployed in the cloud detector, and the result is returned to the edge controller. The controller can adopt corresponding protective measures to protect network security according to the type of attack. It is verified by experiments that NB-BP can betray normal and abnormal traffic more accurately than NB and BP models, while EDL model has higher attack category recognition accuracy than a single deep learning model.