Encrypted Traffic Classification Based on Text Convolution Neural Networks

Traffic classification is of great significance to network management. With the wide application of encryption technology, traditional traffic classification methods are no longer applicable. In recent years, the machine learning-based traffic classification method has become more and more popular. However, this method not only needs complex feature engineering but also is difficult to adapt to different network environments. A traffic classification method based on text convolution neural network is proposed in this paper, which represents traffic data as vectors, then use text convolution neural networks to extract key features for traffic classification. This method is validated on ISCX VPN-non VPN dataset and achieves better classification performance than the previous traffic classification method. For class imbalance problem, a new loss function and an appropriate method of class weight allocation are used in multi-class classification task, which can effectively deal with class imbalance problem.