Fine-grained Traffic Classification based on Functional Separation

Abstract

The most critical and representative issue that exists in today's Internet traffic classification is achieving a high-level of accuracy and completeness. A decade of research on traffic classification has provided various methodologies to investigate the traffic composition in data communication networks. Many variants or combinations of such methodologies have been continuously introduced to improve the classification accuracy and efficiency. However, due to the fast-changing nature of the Internet traffic, it is extremely difficult for any method to achieve 100% accuracy and completeness. This indicates that traffic classification research still has a lot of room for improvement.Another issue in traffic classification is the analysis of traffic classification results. Previous studies have discussed various classification methodologies. Yet, the level of classification details is often subject to identifying protocols or applications in use. The main causes of low accuracy and completeness are new types of network applications and their complex characteristics. In recent years, the majority of traffic classification studies have focused on detecting major applications such as P2P and streaming applications, which occupy most of the Internet traffic. Furthermore, detecting P2P or streaming also has concentrated on classifying main functions that generate a great portion of traffic workload.In this context, this thesis proposes a new traffic classification scheme called a fine-grained traffic classification based on the analysis of existing classification methodologies. Our unique traffic scheme can increase traffic classification accuracy and completeness by reducing the amount of undetected traffic and provide more in-depth classification results for various analyses which are unable to be achieved by current traffic classification schemes and methods. The key to the fine-grained traffic classification is classifying network flows into different functional groups based on origin functions in an application. To achieve this, we also propose functional separation method. By applying this method we are able to detect different types of traffic generated by a single application according to their functionalities. The fine-grained traffic classification based on functional separation will potentially increase the amount of information that can be obtained by traffic classification and lay a cornerstone in the foundation of applying traffic classification in user-behavior analysis.