Hope of Delivery: Extracting User Locations From Mobile Instant Messengers

TL;DR

This paper reveals that delivery status notifications in mobile instant messengers can be exploited as a timing side channel to infer user locations, even in privacy-focused apps, posing significant privacy risks.

Contribution

The study demonstrates the existence and persistence of a timing side channel in popular messengers, and evaluates potential countermeasures to mitigate this privacy vulnerability.

Findings

Timing side channel exists in WhatsApp, Signal, and Threema.

Attack accuracy exceeds 80% within the same city.

Random delays can effectively mitigate the attack.

Abstract

Mobile instant messengers such as WhatsApp use delivery status notifications in order to inform users if a sent message has successfully reached its destination. This is useful and important information for the sender due to the often asynchronous use of the messenger service. However, as we demonstrate in this paper, this standard feature opens up a timing side channel with unexpected consequences for user location privacy. We investigate this threat conceptually and experimentally for three widely spread instant messengers. We validate that this information leak even exists in privacy-friendly messengers such as Signal and Threema. Our results show that, after a training phase, a messenger user can distinguish different locations of the message receiver. Our analyses involving multiple rounds of measurements and evaluations show that the timing side channel persists independent of distances between receiver locations -- the attack works both for receivers in different countries as well as at small scale in one city. For instance, out of three locations within the same city, the sender can determine the correct one with more than 80% accuracy. Thus, messenger users can secretly spy on each others' whereabouts when sending instant messages. As our countermeasure evaluation shows, messenger providers could effectively disable the timing side channel by randomly delaying delivery confirmations within the range of a few seconds. For users themselves, the threat is harder to prevent since there is no option to turn off delivery confirmations.