Network Encrypted Traffic Classification Based on Secondary Voting Enhanced Random Forest

Abstract

Nowadays, traffic classification plays a significant role in network behavior analysis, network planning, network anomaly detection and network traffic model construction. Since the Internet is indubitably moving towards the era of encryption, making traffic classification more and more challenging. In this paper, a novel classification model based on random forest is proposed, and public data set named ISCX VPN-NonVPN is adopted for validation. Compared with the general random forest, the classification process is divided into two parts: two-group test and secondary voting. The accuracy of prediction can be improved by secondary voting. Results prove that our method can achieve averagely 5% higher precision than comparison methods which includes the decision tree method, KNN and general random forest.