The L2 norm of a complex vector in cvx

That error message basically says you have one or more quadratic elements which are not recognizably convex due to being squares. It has nothing to do with whether a matrix has square dimensions.

You haven’t followed DCP rules. Moreover, your expression is non-convex (unless h is all zeros, or N=1), so reformulation to be acceptable to CVX is not possible. If there is a convex approximation suitable for your purposes, that is for you to determine, because it’s your problem, so you presumably know why you’re trying to solve it.