Physical layer authentication in the internet of vehicles through multiple vehicle-based physical attributes prediction

The high security and low complexity of physical layer authentication (PLA) make it a promising complement for complex cryptographic authentication approaches, especially for Internet of Vehicles (IoV) systems constrained by limited computational capabilities and facing increasing security threats due to the broadcast nature of their communications. However, existing PLA schemes exploiting geographical location information to predict channel characteristics for authentication face the challenge of location falsifying/spoofing attacks. Moreover, they either require cryptographic-based initial authentication or cannot be applied to moving vehicle scenarios. To tackle these challenges, we propose a PLA scheme based on the Gaussian process (GP) regression that jointly considers the location and speed attributes of vehicles to enhance the reliability of authentication in the IoV network. Specifically, the historical channel state information attributes together with the location and speed information of transmitters are utilized to establish a mapping and train a GP model to predict the next legitimate location and speed of a transmitter for authentication. First, for vehicle-to-road side unit (RSU) authentication, the trained GP model is stored on RSU and used to authenticate vehicles entering its vicinity by cross-verifying their reported location and speed information with the ones predicted by the model. Next, for vehicle-to-vehicle authentication, we propose an RSU-assisted authentication, where the RSU in the location shared by two vehicles is used to assist in verifying the validity of their reported location and speed information. Finally, for RSU-to-vehicle authentication, we leverage the path loss and angle of arrival of a signal from RSU to estimate and cross-verify its location. We utilize QuaDRiGa, a quasideterministic radio channel generator to generate realistic channels for experimental validations. The results of simulation tests conducted demonstrated that our approach significantly improves authentication performance compared with the existing approaches.