SOFTWARE-DEFINED NETWORK TRAFFIC CLASSIFICATION METHOD BASED ON SPATIO-TEMPORAL ATTENTION

It is necessary that emerging technologies such as 5G and B6G promote the continuous increase of network scale and increase the difficulty of traffic control. SDN can realize centralized control of network traffic. However, the existing methods are difficult to accurately describe the temporal and spatial characteristics of network traffic, and the algorithm convergence speed is low. In response to the above problems, this paper proposes a software-defined network traffic classification method based on spatio-temporal attention. The model was composed of two components, spatial and temporal feature extraction. The spatial component included a channel and a spatial attention module, and the temporal component was composed of a temporal attention module. It was composed of multi-layer two-way GRU stacks, and an innovative reconstruction mechanism integrated time and space characteristics. The simulation results show that this method is significantly better than the existing baseline method in classification performance, the algorithm converges fast, and has better classification performance under different types of traffic.