A New Semi-Supervised Method for Network Traffic Classification Based on X-Means Clustering and Label Propagation

Abstract

Network traffic classification is an essential requirement for network management. Various approaches have been developed for network traffic classification. Traditional approaches such as analysis of port number or payload have some limitations. For example, using port numbers for traffic classification fails if an application uses dynamic port number or applies encryption methods. To address such limitations, modern traffic classification methods employ machine learning techniques. However, machine learning-based traffic classification needs a large labeled data to extract accurate classification model which is expensive and time-consuming. To overcome this issue, we propose a new semi-supervised method for traffic classification based on x-means clustering algorithm and a new label propagation technique. The accuracy of the proposed method tested on Moore's dataset is 0.95 that shows its effectiveness for learning a network traffic classifier using a limited labeled data.