改进移动加密流量分类的方法——数据质量分数OA

The rapid development of mobile internet has led to a surge in demand for classifying en-crypted mobile traffic.Deep learning classification methods rely on data features,but there are differ-ences in the feature quantities of different data,and evenly distributing weights may decrease perfor-mance.To address this issue,we propose a method—Data Quality Score(DQS)to differentiate data and use different weights in the loss function to reduce the interference of low-quality data on model parameters,while enhancing the effect of high-quality data.The effectiveness of this method is verified through experiments on the Mirage-2019 dataset.We first conduct statistical analysis on this dataset to determine feature selection.Then,we build classification models with different neural network struc-tures for experiments and compare their performance with and without DQS method.Results of 5-fold cross-validation indicate that after incorporating the DQS method,the classification performance of dif-ferent network models has been improved without apparent increase in training time.