Multiple attributes based physical layer authentication through propagation scenario identification in the internet of vehicles

Abstract

The Internet of Vehicles (IoV) systems improve road safety through coordination among vehicles, but they are vulnerable to impersonation attack due to their broadcast communication nature. Existing cryptographic authentication approaches are complex for resource-limited IoV devices, which motivated the development of less complex Physical Layer Authentication (PLA) approaches. However, the existing PLA approaches use all channel attributes for all communication scenarios and do not update reference attributes used to authenticate newly observed attributes during authentication, which leads to poor performance when the attributes become insensitive to communication scenarios. To address these limitations, we propose a multiple attributes-based PLA scheme that flexibly selects effective attributes according to current communication scenarios for improved authentication. First, the historical channel attributes including the ricean K-Factor (KF), the Delay Spread (DS), and the Azimuth Spread of Arrival (ASA) together with the communication scenarios (i.e., Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS)) information of transmitters are used to establish a relationship and train a Long Short-Term Memory (LSTM) neural network to identify current communication scenarios. The softmax function is then used to evaluate the scenario identification performance of each attribute and the attributes with maximum softmax values are dynamically selected for authentication. Second, to track and update the reference attributes, we use the historical attributes together with the past location and communication scenarios information of transmitters and train our model to predict future attributes for authentication. Finally, we use the Euclidean distance to measure the similarity between the selected and the predicted attributes to authenticate the transmitters. We validate the effectiveness of our approach and compare its performance with the existing methods through extensive simulations using realistic radio channel characteristics generated from the Quasi Deterministic Radio channel Generator (QuaDRiGa) platform. The results of the evaluation show that our system improves authentication performance and its false alarm and miss detection are respectively 19.19% and 52% lower than the existing approaches.