Towards smart phone traffic classification

The appearance of smart phones and their continuing rapid uptake has large affects on our society in as much as they represent a paradigm shift in the traditional industrial structure. The Telecom market is changing day by day, networks with the complicated and varied traffic have almost reached capacity because of the rapid increase of user and the service releases on smart phones. Therefore, the necessity for smart phone traffic monitoring and analysis has increased. Traffic analysis is an essential element for efficient and reliable networks. In this paper, we propose a new smart phone traffic classification by application method. The proposed method is composed of several consecutive steps: grouping the HTTP User-Agent field, extracting common strings by the LCS algorithm and finally classifying the traffic. In addition, to classify unknown traffic from previous methods, we propose a process that extracts header signatures in grouped information to improve the classification completeness. We achieved about a 90% accuracy rate for the analysis by our proposed method in the target campus network.