Non-asymptotic analysis of secrecy capacity in massive MIMO system

In this paper, we consider a massive MIMO wiretap system where the transmitter, the receiver and the eavesdropper are equipped with a large number of antennas. Being different from the previous works using asymptotic random matrix theory, our analysis relies on the concentration measure of non-asymptotic random matrix theory which allows us to obtain tight bounds for secrecy capacity of massive MIMO system with finite antenna number. The analytical and simulation results reveal the following, in the massive MIMO system employing equal power allocation at each transmit antenna: 1) the secrecy capacity falls within a bounds with a probability growing exponentially with the number of transmit antenna, while the ergodic secrecy capacity falls within a deterministic bounds; 2) the gap between the upper and the lower bound on secrecy rate is proportional to the square root of the SNR at legitimate receiver and the SNR at eavesdropper, respectively; 3) when the entry of legitimate channel matrix and eavesdropping channel matrix satisfies Gaussian distribution, the gap between the upper and the lower bound on secrecy rate is a linear reciprocal function of the number of transmit antennas.