Improving traffic classification and policing at application layer

The common process of classifying network traffic resorting to a set of IP header fields and well-known communication ports is highly fallible as some applications try to hide their true nature by, for instance, using dynamic, non default ports. In this paper, we argue and demonstrate that application layer inspection is a possible and convenient approach to derive the correct application protocol. This detection and classification process is crucial to allow an efficient control of traffic entering the network. Taking pfSense as a case study, we extend its current layer 3 and 4 classification scheme with layer 7 (L7) capabilities, providing a powerful solution to control traffic based on application patterns. We propose the concept and use of L7 containers so that a user can easily create a set of rules for inspection, which will drive lower-level traffic control. In addition, we propose and implement a mechanism to create automatically useful application inspection scenarios.