Physical Layer Authentication for Mobile Systems with Time-Varying Carrier Frequency Offsets

A novel physical layer authentication scheme is proposed in this paper by exploiting the time-varying carrier frequency offset (CFO) associated with each pair of wireless communications devices. In realistic scenarios, radio frequency oscillators in each transmitter-and-receiver pair always present device-dependent biases to the nominal oscillating frequency. The combination of these biases and mobility-induced Doppler shift, characterized as a time-varying CFO, can be used as a radiometric signature for wireless device authentication. In the proposed authentication scheme, the variable CFO values at different communication times are first estimated. Kalman filtering is then employed to predict the current value by tracking the past CFO variation, which is modeled as an autoregressive random process. To achieve the proposed authentication, the current CFO estimate is compared with the Kalman predicted CFO using hypothesis testing to determine whether the signal has followed a consistent CFO pattern. An adaptive CFO variation threshold is derived for device discrimination according to the signal-to-noise ratio and the Kalman prediction error. In addition, a software-defined radio (SDR) based prototype platform has been developed to validate the feasibility of using CFO for authentication. Simulation results further confirm the effectiveness of the proposed scheme in multipath fading channels.

I. Introduction

Security provisioning has become a fundamental challenge in wireless communications due to the open nature of the shared air space as a communications medium, where the transmitted signal is broadcasted to any receiver within its coverage [1]–[3]. Traditional wireless security schemes, relying on data encryption and authentication at the higher layers of the protocol stack, do not directly address physical layer vulnerabilities and security threats from the open air interface. In addition, conventional wireless security measures often lead to excessive communication latency, high power consumption and decreased system capacity due to the imposed computational load and signaling overhead [4], [5].