Research On Encrypted Traffic Classification Method Based On Deep Learning

With the rapid development of the Internet and the continuous enhancement of users' information security awareness,encryption of network traffic has become a basic method of data transmission.The rapid growth of encrypted traffic in the network and the increasingly complex encryption algorithms make it difficult for traditional classification methods to meet the current network environment.Deep learning has demonstrated good classification performance on encrypted traffic classification tasks due to its excellent data representation capabilities.Most of the existing encrypted traffic classification methods use a large amount of labeled data for training,but labeling encrypted traffic requires experts to analyze the collected data,which not only bears high labor costs,but also risks leaking user privacy.At the same time,the existing deep learning-based encrypted traffic classification models are complex in structure and require a large amount of computation,and they often perform poorly in practical network scenarios.Based on the above problems encountered by deep learning in the field of encrypted traffic classification,this paper proposes the following work:Aiming at the problem that label data is not easy to obtain in encrypted traffic,a semisupervised method of encrypted traffic classification based on convolutional neural network is proposed.This method can make full use of a large amount of unlabeled data and a small amount of labeled data for learning.Firstly,each sample needs to be encoded in the form of picture,and the traffic statistical characteristics are extracted.Model training is divided into two stages,unsupervised pre-training and re-training.During the unsupervised pre-training process,the traffic information labels obtained by combining statistical information and Kmeans clustering algorithm will be used as labels for the unlabeled traffic graph.After the unsupervised pre-training is completed,a small number of labeled traffic graphs are used for retraining through model parameter migration to obtain the final classification model.Aiming at the problems of large amount of computation and complex network in the encrypted traffic classification task,an encrypted traffic classification method based on spatiotemporal features and self-attention is proposed.This method extracts packet sequences from raw traffic data,combining convolutional neural network and improved Transformer selfattention encoder,and constructs a lightweight network model.The model reduces the computational complexity while ensuring the model classification accuracy.The availability of the model is proved by classification results and comparative experiments.The method proposed above is verified on the public encrypted traffic dataset ISCX VPNNon VPN.The experimental results show that the method proposed in this paper has better performance than the baseline method in encrypted traffic classification.