Smart device fingerprinting based on webpage loading

Abstract

Detecting devices connected to a network has become of serious importance for the network. Different devices differ in CPU scheduler, screen resolution and clock frequency, resulting in different performances when loading the same webpage. In this paper, we present a content-agnostic device identification method, a technique which decomposes webpage loading time and loads as the features to identify physical devices. This proposed method can deal with various types of devices such as mobiles, laptops, and other smart devices. We conduct experiments to evaluate the performance of the proposed method with real-world traffic. The experiment results demonstrate that the proposed method can accurately identify the types of devices from encrypted traffic and the recognition rate can reach 98.4%. To demonstrate the scalability of the method, we heuristically applied it to website identification and found that it has better effects than existing methods.