Location-leaking through Network Traffic in Mobile Augmented Reality Applications

Abstract

Mobile Augmented Reality (AR) applications allow the user to interact with virtual objects positioned within the real world via a smart phone, tablet or smart glasses. As the popularity of these applications grows, recent researchers have identified several security and privacy issues pertaining to the collection and storage of sensitive data from device sensors. Location-based AR applications typically not only collect user location data, but transmit it to a remote server in order to download nearby virtual content. In this paper we show that the pattern of network traffic generated by this process alone can be used to infer the user's location. We demonstrate a side-channel attack against a widely available Mobile AR application inspired by Website Fingerprinting methods. Through the strategic placement of virtual content and prerecording of the network traffic produced by interacting with this content, we are able to identify the location of a user within the target area with an accuracy of 94%. This finding reveals a previously unexplored vulnerability in the implementation of Mobile AR applications and we offer several recommendations to mitigate this threat.