Internet of Profiling - Traffic, Users and Applications

Recent years have witnessed a significant increase in monitoring network traffic in order to profile user behaviour and to provide better service. This paper will provide a review of these efforts, highlighting the benefits brought by traffic profiling, particularly in relation to providing a better user experience and higher quality of service. The discussion will focus on three headings: identifying and profiling applications based on statistical analysis of traffic, identifying users and anomaly detection based on network interaction, and providing fairness in a heterogeneous user environment. Profiling applications is a challenging task in the context of encryption and tunnelling, but allows better provision of network resources, in line with the needs of each application, from email to video streaming. Identifying users may raise concerns in terms of privacy, but the primary aim is not to single them out but to cater for their needs at an aggregate level, both in terms of dealing with significant variations as well as potentially acting as a first line of defence when anomalies are detected. Finally, while globally the behaviour of users may appear similar, there is significant variation in terms of the demand, usage, and expectations of each user; ensuring fairness in such a diverse environment requires acknowledging the user requirements and accommodating them against a heterogeneous environment in terms of provision and demand. The presentation will draw from a number of research studies undertaken over the recent years in the above areas, both across the research community as well as at Plymouth University, and discuss how the findings impact on the wider user community.