<i>SoundUAV</i>

Abstract

Drones are gaining a lot of traction in a wide spectrum of applications. This popularity makes them attractive attack surfaces, which necessitates the need for ensuring their security. Specifically, in the case of drone delivery, an attacker drone may impersonate the legitimate one in order to steal packages, which makes drone authentication important. Recent efforts have pushed to incorporate digital certificates as an authenticator for drones. However, such software-based techniques are often compromised and can be launched on a large scale, making them a bigger threat. To this end, we propose SoundUAV as an additional factor of authentication that leverages differences in the acoustic noise characteristics of drones to fingerprint them. These differences are caused due to manufacturing defects in their motors, making them hard to replicate. Moreover, SoundUAV does not require any hardware modifications to existing drones as they leverage the freely available sound information, and it is also robust to large-scale attacks as the attacks involve hardware alterations. To test the feasibility of SoundUAV, we evaluate it on 54 motors, and 11 drones of the same make and model, and report fingerprinting accuracy of 99.48%.