CCS: A Cross-Plane Collaboration Strategy to Defend Against LDoS Attacks in SDN

Software-Defined Networking (SDN) actualizes the separation of control and forwarding, innovates network functionality with a logically centralized controller, and facilitates network-wide collaboration. Contemporary SDN infrastructure exposes potential bottlenecks which are prone to engaging low-rate denial of service (LDoS) attacks. Currently, a great deal of detection methods are deployed in the controller, and the controller needs to poll the switch frequently, which brings heavy load to the controller and the southbound link. According to the analysis of existing researches, we focused on the how to decrease the frequent polling of the controller and improve the detection rate. In this paper, we adopted the idea of cross-plane collaboration and proposed a two-phase detection framework, which carried out the lightweight detection method in the data plane and the in-depth detection based on Bayesian voting mechanism in the control plane. Once LDoS attacks are detected, the controller recalculates routes for the bottleneck nodes using the optimized Dijkstra algorithm to complete mitigation. Theoretical analyses and extensive experiments are conducted to validate the performance of our proposed method. Test results show that our method outperforms other traditional methods in terms of the detection rate of 99.1%, the detection delay of 1.3s and the communication overhead of 1068 Byte/s, the average CPU utilization of controller remains at approximately 3.5%. The proposed method takes a step forward to enhance the security of SDN.