Intelligent Network Client Profiler

Abstract — Peer2Peer traffic already accounts for a large share of the overall internet traffic. Future solutions will need to manage all the available resources in order to charge users using fair rules according to their communication profile. Obtaining information about the behavior of Internet traffic is therefore fundamental to the management, monitoring and operation activities, such as the identification of applications and protocols that customers use. However, the main obstacle to this identification is the lack of scalability for monitoring network devices. In particular, they can analyze all the network packets for this purpose. This task is extremely demanding and almost impossible to accomplish rapidly in large networks (because usually there is a number in the hundreds or thousands of customers). Furthermore, we expect such networks to become even larger, as on the internet of things all devices (sensors, appliances, etc) will be publicly connected to the internet. As such, traffic sampling strategies have been proposed to overcome this major problem of scale. This paper presents different works in the area of monitoring traffic for user profiling and security purposes. It proposes as well a solution that uses selective filtering techniques combined with an engine traffic DPI to identify applications and protocols that customers use most frequently. Thus it becomes possible to get ISPs to optimize their network in a scalable and intelligent manner, imposing security measures in order to enforce network usage according to client profiles.