Physical Layer Encryption for UAV-to-Ground Communications

Ensuring secure and reliable wireless communication is crucial for Unmanned Aerial Vehicle (UAV) applications. Most of the prior works on secure UAV-to-Ground (U2G) communications focus on trajectory and/or power optimization to ensure that the desired U2G channel is stronger than an eavesdropping channel. In this paper we propose a novel physical layer encryption method that performs symbol and/or constellation hiding for secure U2G communications. Unlike prior works on symbol and/or constellation hiding which aimed at specific detection algorithms by adversaries, our method exploits the secrecy inherent in the reciprocal channel between an UAV and a desired ground station (GS), and is hence in principle robust against any eavesdropping attack algorithms including deep machine learning. Given a pair of estimated reciprocal channel vectors (ERCVs) with a limited dimension at UAV and GS respectively, our method first uses a continuous encryption function (CEF) to transform the two ERCVs at UAV and GS respectively into two sequences of quasi-continuous pseudo-random numbers (QCPRNs) of any desired dimension. Robust to a range of statistical distributions of ERCVs, these QCPRNs follow approximately a known statistical distribution and hence can be further transformed into two sequences of uniformly dis-tributed (UD) QCPRNs. The UD-QCPRNs generated at UAV are superimposed by UAV in a modulo fashion onto its transmitted symbols, and the UD-QCPRNs generated at GS are used for decryption at GS. This paper also studies the impact of the difference between the two ERCVs along with other noises on the performance of the desired U2G channel.