1: \begin{abstract}

2: %This paper considers the secrecy capacity of a multiple-input multiple-output

3: %(MIMO) Gaussian wiretap channel (WTC), which involves solving a difference

4: %of convex functions program known to be non-convex for the non-degraded

5: %case.

6: To achieve perfect secrecy in a multiple-input multiple-output

7: (MIMO) Gaussian wiretap channel (WTC), we need to find its secrecy capacity and  optimal signaling, which involves solving a difference

8: of convex functions program known to be non-convex for the non-degraded

9: case.

10:  To deal with this, a class of existing solutions have been developed

11: but only local optimality is guaranteed by standard convergence analysis.

12: Interestingly, our extensive numerical experiments have shown that

13: these local optimization methods \emph{indeed achieve global optimality}.

14: In this paper, we provide an analytical proof for this observation.

15: To achieve this, we show that the Karush-Kuhn-Tucker (KKT) conditions

16: of the secrecy rate maximization problem admit a unique solution for \emph{both degraded and non-degraded cases}.

17: Motivated by this, we also propose a low-complexity algorithm to find

18: a stationary point. Numerical results are presented to verify the

19: theoretical analysis.

20: \end{abstract}

21: