Multidimensional Secret Key Agreement with Tensor-Decomposition-Based mmWave MIMO Channel Estimation.

Considering the rich parameters used in characterizing multiple-input-multiple-output (MIMO) channels in the millimeter wave (mmWave) band, a highly efficient approach is proposed for secret key agreement exploiting variables in multiple dimensions as the sources of common randomness for advantage distillation (AD) between the legitimate users. In order to simultaneously acquire multiple channel state information (CSI) parameters so that the AD efficiency is improved, channel estimation based on a high-dimensional tensor model for mmWave MIMO channels is considered. Specifically, the mmWave MIMO channel is modeled as a high-dimensional tensor that spans the space-time-frequency dimensions of the channel. Bi-directional channel estimation is then conducted through tensor decomposition to extract the CSI parameters at both legitimate users. Canonical Polyadic (CP) decomposition method, which has uniqueness guarantee for tensor decomposition results, is employed. A key agreement protocol based on bi-directional channel identification is designed accordingly, and the corresponding key rate performance is evaluated. Numerical results validate that the proposed method, by simultaneously acquiring multiple common randomness parameters on both legitimate parties, achieves higher secret key rate compared with conventional channel identification methods that process one parameter at a time.