User Profiling Based on Application-Level Using Network Metadata

There is an increasing interest to identify users and behaviour profiling from network traffic metadata for traffic engineering and security monitoring. Network security administrators and internet service providers need to create the user behaviour traffic profile to make an informed decision about policing, traffic management, and investigate the different network security perspectives. Additionally, the analysis of network traffic metadata and extraction of feature sets to understand trends in application usage can be significant in terms of identifying and profiling the user by representing the user's activity. However, user identification and behaviour profiling in real-time network management remains a challenge, as the behaviour and underline interaction of network applications are permanently changing. In parallel, user behaviour is also changing and adapting, as the online interaction environment changes. Also, the challenge is how to adequately describe the user activity among generic network traffic in terms of identifying the user and his changing behaviour over time. In this paper, we propose a novel mechanism for user identification and behaviour profiling and analysing individual usage per application. The research considered the application-level flow sessions identified based on Domain Name System filtering criteria and timing resolution bins (24-hour timing bins) leading to an extended set of features. Validation of the module was conducted by collecting Net Flow records for a 60 days from 23 users. A gradient boosting supervised machine learning algorithm was leveraged for modelling user identification based upon the selected features. The proposed method yields an accuracy for identifying a user based on the proposed features up to 74%.