No abstract

Network traffic classification is a vital task for service operators, network engineers, and security specialists to manage network traffic, design networks, and detect threats. Identifying the type/name of applications that generate traffic is a challenging task as encrypting traffic becomes the norm for Internet communication. Therefore, relying on conventional techniques such as deep packet inspection (DPI) or port numbers is not efficient anymore. This paper proposes a novel flow statistical-based set of features that may be used for classifying applications by leveraging machine learning algorithms to yield high accuracy in identifying the type of applications that generate the traffic. The proposed features compute different timings between packets and flows. This work utilises tcptrace to extract features based on traffic burstiness and periods of inactivity (idle time) for the analysed traffic, followed by the C5.0 algorithm for determining the applications that generated it. The evaluation tests performed on a set of real, uncontrolled traffic, indicated that the method has an accuracy of 79% in identifying the correct network application.

Classifying Internet traffic into applications is vital to many areas, from quality of service (QoS) provisioning, to network management and security. The task is challenging as network applications are rather dynamic in nature, tend to use a web front-end and are typically encrypted, rendering traditional port-based and deep packet inspection (DPI) method unusable. Recent classification studies proposed two alternatives: using the statistical properties of traffic or inferring the behavioural patterns of network applications, both aiming to describe the activity within and among network flows in order to understand application usage and behaviour. The aim of this paper is to propose and investigate a novel feature to define application behaviour as seen through the generated network traffic by considering the timing and pattern of user events during application sessions, leading to an extended traffic feature set based on burstiness. The selected features were further used to train and test a supervised C5.0 machine learning classifier and led to a better characterization of network applications, with a traffic classification accuracy ranging between 90-98%.