In this paper, we design a secure centralized detection network where each sensor employs an amplify-forward strategy to transmit its observation to the fusion center (FC), along with artificial noise using multiple transmitting-antennas in order to confuse the eavesdropper (Eve). We maximize KL Divergence at the FC in the presence of a constraint on the Eve's KL Divergence, and a total energy constraint on each sensor. Since this problem is non-convex and intractable, we propose a two-stage algorithm that provides an approximate solution to the desired problem. Numerical results are presented to illustrate the efficiency of our design algorithm.