Physical-Layer Authentication Over Spatially Correlated MIMO Channels

We propose a novel physical-layer authentication (PLA) method that exploits spatio-temporal correlation features of multiple-input multiple-output (MIMO) channels. In particular, in the proposed method, the correlation due to the spatial structure and aging of the channel between two legitimate nodes (Alice and Bob), captured in the corresponding covariance matrix, is incorporated into the Mahalanobis distance between consecutive channel estimates and utilized to detect the attack by an impersonator. The authentication decision is carried out via hypothesis testing, with distributions and threshold derived analytically. Computer simulations demonstrate that, for a given false alarm probability, the proposed method can detect an attacker with a higher probability than a conventional method based on Euclidean distances between channels.