Research on Encrypted Traffic Detection Technology Based on Fusion Attention Mechanism

Network encrypted traffic is widely used in network security and management, which requires obtaining differentiated traffic representations from invisible content and unbalanced network traffic to achieve accurate classification, which is a great challenge. For this purpose, this article proposes an f model based on fusion attention mechanism. Convolutional neural networks are used to extract spatial features of traffic, and the spatiotemporal correlation features between encrypted network traffic are analyzed using the BiLSTM (Bidirectional Long Short-Term Memory Network) method. SE (Squeeze-and-Excitation Attention) and CBMA (Convolutional Block Attention Module) are then used to reassign and weight the spatial and channel dimensions. Using the USTC-TFC2016 datasets, the network flow was split from the traffic level to the session level using SplitCap. Afterwards, generate a fixed byte image. Compared with a single CNN (Convolutional Neural Network) and CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory Network) structure, the proposed encrypted traffic model based on fusion attention mechanism exhibits better classification performance in this paper.