Application traffic identification based on remote subnet grouping

Recently, the number of Internet applications available for use on both desktop computers and smartphones has rapidly increased. The Internet traffic generated by these applications has increased significantly as well. Network operators are required to be aware of the status of managed networks in terms of application usage. However, the application traffic observed in a network is dependent on users and the geographical location of the network. As a result, selecting applications that are expected to appear frequently in the network is required as the initial step in the generation of ground-truth traffic and extraction of classifiers. In this paper, we propose a traffic identification methodology for the initial step of the classifier generation procedure. The proposed approach is based on remote subnet grouping, which refers to the collection of the same (or similar) application traffic. We also validate the proposed methodology in terms of completeness, feasibility, and accuracy by using the traffic in a campus network.