Malicious Controller Detection Mechanism based on Graph Convolutional Neural Network.

Software Defined Networking (SDN) has rapidly developed in recent years as an emerging network technology. It achieves decoupling between the control and data planes, where the control layer is responsible for managing the network and maintaining a global network topology view, while the data layer is responsible for data forwarding. The decoupling of the control and data planes simplifies network management and enables network programmability. In SDN, the control layer is the core of the entire network, not only responsible for managing the forwarding devices in the data layer, but also providing accurate network status information to the application layer. Therefore, the security and reliability of the SDN controller are crucial. We propose a controller detection method. By obtaining feature information of the controllers in the network through in-band telemetry, and after multiple rounds of model training, the mechanism can detect malicious controllers. Based on the constructed network simulation environment, it can effectively detect malicious controllers in the network and enhance the security and reliability of the network.