Optimal jamming with codewords

In this paper for a noise forwarding (NF) scheme, the joint design of the optimal covariance matrices of a source's and a helper's signals that maximize the secrecy rate is studied. Under general power constraints, it is shown that the non-convex design problem can be converted into a quasi-convex optimization problem composed of convex subproblems. The properties of the optimal covariance matrices are examined. Under the individual power constraint, beamforming is shown to be optimal, and the design problem is reduced to a problem with two real variables. The performances of the NF scheme and the artificial noise (AN) scheme are compared. The design of a hybrid scheme consisting of the NF and AN techniques is also addressed.