Research On Key Algorithms For Class-imbalanced Network Encryption Traffic Classification

Network encryption traffic classification is of great significance to network security management and control.Traditional encrypted traffic classification algorithms are often proposed based on the assumption of data balance.However,in actual network scenarios,traffic data generally has class imbalance characteristics,which leads to a significant decrease in the performance of traditional classification algorithms.Classification is not good.In view of the above problems,this paper analyzes and studies the class imbalance characteristics of network encrypted traffic,and proposes two traffic classification algorithms to solve the class imbalance problem of network encrypted traffic from the data and classification algorithm levels.The main contents include the following three aspects:(1)The apparent imbalance and inherent imbalance characteristics of network encrypted traffic are studied and analyzed,the principles and applications of current typical traffic classification algorithms and imbalanced learning algorithms are introduced,and the influence mechanism of class imbalance on classification algorithms and the current state of the algorithm are analyzed and verified through experiments.There are deficiencies in the existence of imbalanced learning algorithms.(2)In view of the limited improvement of the classification performance of the classification model by the existing resampling methods and the fact that it is easy to lead to overfitting,this paper proposes a class-imbalanced encrypted traffic classification algorithm based on improved generative adversarial networks from the data level.The algorithm adds conditional constraints on the basis of the traditional generative adversarial network,and introduces the Wasserstein distance metric as an indicator to measure the distance between the real and generated data distributions.At the same time,gradient penalty is added,and the Image CWGAN-GP model is designed to generate a minority class.sample.Finally,a classification model based on convolutional neural network CNN is used to train and classify the expanded balanced dataset.The experimental results show that the algorithm significantly improves the classification performance of the classification model on class-imbalanced traffic data.(3)Aiming at the problem that the classification algorithm based on the traditional convolutional neural network model has poor classification performance in the face of class imbalanced encrypted traffic,this paper proposes an improved convolutional neural network based on multi-scale feature fusion and attention mechanism.Classification algorithm.The proposed algorithm realizes multi-scale feature fusion by adding an Inception module on the basis of the traditional convolutional neural network model,so that the model can better learn the feature distribution of the samples,and alleviates the difficulty of the model due to the scarcity of the number of minority samples.The problem of learning the feature distribution well.At the same time,the attention module is introduced to make the model pay attention to more discriminative features and enhance the representation ability of the model.The experimental results show that the algorithm can effectively alleviate the impact of class imbalance on classification and improve the classification performance of the classification model on the overall and minority classes.