摘要：

基于浅层机器学习的加密流量识别方法准确率偏低,在特征提取和选择方面耗时耗力。为此,提出一种基于堆栈式自动编码器(SAE)的加密流量识别方法。该方法利用SAE的无监督特性及在数据降维等方面的优势,结合多层感知机(MLP)的有监督分类学习,实现对加密应用流量的准确识别。考虑到样本数据集的类别不平衡性对分类精度的影响,采用SMOTE过抽样方法对不平衡数据集进行处理。实验结果表明,该方法各项性能指标均优于MLP加密流量识别方法,识别精确度和召回率以及F1-Score均可达到99%。

关键词: 加密流量识别, 深度学习, 堆栈式自动编码器, 流量分类, 多层感知机, 卷积神经网络

Abstract: To solve the problem that encrypted traffic identification methods based on machine learning are low in accuracy and time-consuming and costly in feature extraction and selection,this paper proposes a Stacked Autoencoder (SAE)-based encrypted traffic identification method.The method utilizes the unsupervised characteristics of SAE and its advantages in dimensional reduction,combined with supervised classification learning of Multilayer Perceptron (MLP) to achieve accurate identification of encrypted application traffic.Considering the influence of the class imbalance of the sample dataset on the classifier performance,the unbalanced dataset is processed by the SMOTE oversampling method.Experimental results show that,the performance indicators of this method are higher than the MLP encrypted traffic identification method,and the precision,recall,and F1-Score can reach 99%.

Key words: encrypted traffic identification, deep learning, Stacked Autoencoder (SAE), traffic classification, Multilayer Perceptron (MLP), Convolutional Neural Network (CNN)

中图分类号: 

• TP302