Polarization Fingerprint-Based LoRaWAN Physical Layer Authentication

Currently, radio frequency fingerprint (RFF), which describes the physical layer features of wireless signals in time-frequency domain, has been intensively studied. However, the research of polarization fingerprint (PF), which describes the physical layer features of wireless signals in the polarization domain, has just started. In this paper, PF is deeply studied, and a PF-based LoRaWAN physical layer authentication solution is designed under the actual application scenario. Firstly, the physical layer features in polarization are analyzed from the full path of polarization formation, propagation and reception, and the mathematical model of PF is constructed. There exist two main properties in PF: frequency characteristic and spatial characteristic. The spatial characteristic is the unique property of PF compared with RFF, which significantly improves the fingerprint discrimination of the same type of device. Subsequently, the solution has two authentication mechanisms according to different deployment strategies of LoRa devices: polarization fingerprint identification (PFI) based authentication and PF tracking-based authentication. PFI is implemented based on ensemble convolutional neural network (CNN), which assigns different weights to parts of PF with different discrimination to improve identification accuracy. PF tracking-based authentication solves the impact of LoRa device movement on authentication. Finally, the authentication performance and robustness of the solution are evaluated by experiments, and the ability of the solution to deal with typical attacks is analyzed. In summary, the PF-based LoRaWAN physical layer authentication solution provides considerable application potential.