GyrosFinger: Fingerprinting Drones for Location Tracking based on the Outputs of MEMS Gyroscopes

Abstract

Drones are widely used for various purposes such as delivery, aerial photography, and surveillance. Considering the increasing drone-related services, tracking the locations of drones can cause security threats such as escaping from drone surveillance, disturbing drone-related services, and capturing drones. For wirelessly monitoring the status of drones, telemetry is used, and this status information contains various data such as latitude and longitude, calibrated sensor outputs, and sensor offsets. Because most of the telemetry implementation supports neither authentication nor encryption, an attacker can obtain the status information of the drones by using an appropriate wireless communication device such as software-defined radio. While the attacker knows the locations of the drones from the status information, this information is not sufficient for tracking drones because the status information does not include any identity information that can bind the identity of the drone with its location. In this article, we propose a fingerprinting method for drones in motion for the binding of the identity of the drone with its location. Our fingerprinting method is based on the sensor outputs included in the status information, i.e., the offsets of micro-electro mechanical systems (MEMS) gyroscope, an essential sensor for maintaining the attitude of drones. We found that the offsets of MEMS gyroscopes are different from each other because of manufacturing mismatches, and the offsets of five drones obtained through their telemetry are distinguishable and constant during their flights. To evaluate the performance of our fingerprinting method on a larger scale, we collected the offsets from 70 stand-alone MEMS gyroscopes to generate fingerprints. Our experimental results show that, when using the offsets of three and two axes calculated from 128 samples of the raw outputs per axis as fingerprints, the F-scores of the proposed method reach 98.78% and 94.47%, respectively. The offsets collected after a month are also fingerprinted with F-scores of 96.58% and 78.45% under the same condition, respectively. The proposed fingerprinting method is effective, robust, and persistent. Additionally, unless the MEMS gyroscope is not replaced, our fingerprinting method can be used for drone tracking even when the target drones are flying.