Guaranteeing Secrecy using Artificial Noise with Quantized Channel Feedback
Shih-Chun Lin,Tsung-Hui Chang,Yan-Lang Liang,Yao-Win Peter Hong,Chong-Yung Chi
2009-01-01
Abstract:The impact of quantized channel direction feedback on the achievable secrecy rate is studied for multiple-antenna wiretap channels. Without knowledge of the eavesdropper’s channel, we consider the scheme proposed by Goel and Negi where artificial noise (AN) is imposed in the null space of the legitimate receiver’s channel to disrupt the eavesdropper’s reception and, thus, guarantee secrecy. It has been shown that the secrecy rate can be made arbitrarily large by increasing the transmission power when perfect knowledge of the legitimate receiver’s channel direction information (CDI) is available at the transmitter. However, in practice, this is difficult to achieve due to rate-limitations on the feedback channel. When only quantized CDI is available at the transmitter, the AN that is only intended to disrupt the eavesdropper’s reception may leak into the legitimate receiver’s channel, causing serious loss in the secrecy rate. Specifically, with quantized CDI, we show that the achievable secrecy rate will now be upper-bounded by a constant and the optimal power allocation between signal and AN must now be more conservative to take into consideration the effects of AN leakage. To overcome the significant rate loss compared to the perfect CDI case, we show that the number of feedback bits must scale at least logarithmically with the transmission power to maintain a constant rate loss. Moreover, we also show that additional channel quality information at the transmitter gains little at the high power regime. These theoretical claims are verified by computer simulations. I. I NTRODUCTION Physical-layer secrecy was first introduced by Wyner [1] under the notion of wiretap channels that typically consist of a transmitter, a legitimate receiver, and an eavesdropper. Recent studies of physical-layer secrecy in the wireless channel have shown that, under perfect secrecy constraints, a non-zero secrecy capacity can always be achieved in fading environments [2] [3]. With