SatPrint: Satellite Link Fingerprinting

Detecting spoofing attacks on a satellite infrastructure is a challenging task, due to the wide coverage, the low received power from the satellite beams and finally the opportunistic nature of radio broadcasting. Although message authentication can be implemented at several communication layers, only a few solutions have been provided at the physical layer---this one exposing features that are invaluable for authentication purposes. Currently available solutions provide physical-layer authentication of the transmitter by combining deep learning and physical-layer features, thus requiring a long and computationally-intensive training process for any new transmitter joining the network. In this work, we propose SatPrint, a solution capable of detecting satellite spoofing attacks by fingerprinting the noise fading process associated with the satellite communication channel. Indeed, the fading of a satellite link is different from the one of a terrestrial link---used very often to launch spoofing attacks---thus allowing one to discriminate between the two. SatPrint does not require retraining when new transducers join the network, and does not rely on hardware impairments of both the transmitter and the receiver. SatPrint has been tested with real satellite and spoofed terrestrial radio measurements, under several different scenario configurations. We prove that SatPrint can effectively discriminate between a satellite transmitter and a fake terrestrial one, with an accuracy greater than 0.99 for all the considered configurations.