Towards fine-grained traffic classification for web applications

Abstract

Web applications, such as video streaming, map services and office applications, have become very popular due to the advances of web technology. Traditional traffic classification methods based on port numbers and payload signatures barely work because the applications run on the same port numbers (usually port 80 and 443) and the payloads are usually encrypted. Furthermore, a web application may provide multiple functions, and the traffic from them has diverse characteristics. In this work, we use statistical features from application messages to characterize the traffic from individual functions of web applications, and perform fine-grained classification to identify the application functions. The experimental results show the classification can achieve high accuracy up to 98.30% for the interaction functions and 92.72% for the download functions.