Exploring the Physical-layer Identification of GSM Devices

Abstract

In this work, we study the physical-layer identification of GSM devices. For our exploration, we build an ad-hoc acquisition setup that collects GSM signals during voice calls. We collect signals from a population of 18 mobile devices and build fingerprints by considering both the transient and the data parts of the acquired signals. Our results show that devices of different models and manufacturers can be identified with high accuracy (0% identification error) by exploiting transient-based fingerprints. Same model and manufacturer devices could also be identified by using transient-based fingerprints: we find an identification error between 0 and 8% depending on the considered device set. We also find that the built transient-based fingerprints are sensitive to the device transmission power, but only partially to the device position with respect to our acquisition setup antenna. This possibly enables defensive (e.g., access control) applications. Although with less accuracy with respect to transient-based fingerprints, data-based fingerprints could also be used to identify same model and manufacturer devices. However, these seem to be sensitive to the device position.