An Improved Two-Way Training for Discriminatory Channel Estimation via Semiblind Approach

This paper studies the discriminatory channel estimation (DCE) performance between a legitimate receiver (LR) and an unauthorized receiver (UR) in the multiple-input multiple-output (MIMO) wireless systems. DCE is a recently developed concept that intentionally degrades the channel estimation at the UR so as to minimize the probability of confidential information being eavesdropped by the UR. Usually, the existing DCE scheme is based on the linear minimum mean square error (LMMSE) method with two-way training. In this paper, we propose a new two-way training for DCE based on semiblind approach, e.g., the whitening-rotation (WR)-based channel estimator. To characterize the DCE performance, we derive the closed-form of the normalized mean squared error (NMSE) to the channel estimation at both the LR and the UR. Simulation results show that the proposed two-way training achieves higher performance compared to the two-way training designs in the literature.