Assessing the Feasibility of RF Fingerprinting for Security in Unmanned Aerial Vehicles

Abstract

The wireless network of consumer drones is particularly vulnerable to remote attacks due to the weak encryption scheme involving the exchange of a Global Unique Identifier (GUID) between transceiver pairs using the binding process, thus exposing the technology to a host of attack vectors such as data spoofing and malicious authentication, among others, leading to security breaches that threaten the prospects of the consumer drone. This study assesses the feasibility of RF fingerprinting as a complementary layer of security devoid of cryptography in the wireless network of unmanned aerial vehicles for enhanced resilience. We evaluate the feature performance of the toy-grade and the universal-grade drone RC transmitters to discern the prospects for device identification in inexpensive, low-end device and the high-end device. Instantaneous amplitude and phase features extracted from the transient phase of time-domain signals acquired off-the-air in the near-field show a high recognition rate in a support vector machine and k-Nearest Neighbour, suggestive of device classification in unmanned aerial vehicle RF hardware, irrespective of built quality.