Impact of estimated CSI quantization on secrecy rate loss in pilot-aided MIMO systems

Abstract

In this paper, we investigate the system performance from the physical layer security provision under imperfect channel state information (CSI). In a classical transmitter (Alice)-legitimate receiver (Bob)- eavesdropper (Eve) model, we introduce artificial noise (AN) to disturb Eve's reception, as Eve's CSI is unknown to Alice. For designing the transmit beamforming vector of the information signal and precoding matrix of AN, Bob feeds back the quantized CSI estimation to Alice. Due to the effects of imperfect CSI at Alice, the secrecy system performance is upper bounded at high SNRs. In order to overcome the problem, by utilizing our derived upper bound on the secrecy rate loss, we put forward a scaled feedback strategy for the secrecy system. By employing the proposed strategy, the secrecy rate increases with transmit power despite that Alice can only obtain imperfect Bob's CSI, and the secrecy rate loss between perfect CSI and imperfect CSI can be controlled within a certain gap. Computer simulations are provided to verify our derived results.