Effective Secrecy is defined by
$D(P_{MZ^n} || P_M Q_{Z^n}) \leq \delta_n$, for some abitary pdf $Q_{Z^n}$.
In the paper "Effective Secrecy: Reliability, Confusion and Stealth" it is written like
$D(P_{MZ^n} || P_M Q_{Z^n}) = \underbrace{I(M;Z^n)}_{\text{Non-Confusion}} + \underbrace{D(P_{Z^n} || Q_{Z^n})}_{\text{Non-Setath}}$
there $I(M;Z^n)$ is the "non-confusion" and $D(P_{Z^n} || Q_{Z^n})$ is the measure of "non-stealth" part and $Q_{Z^n}$ is the distribution that the eavesdropper expects to observe then the source is not comunicating useful messages.
The setup looks like this
M -> Encoder -> X^n -> Q^n_YZ|X -> Y^n -> Decoder -> ^M
-> I(M;Z^n), D(P_{Z^n}||Q^n_Z)
I'm not sure, how to interpret the Non-Stealth part. Has someone a meaning for me? I think, I'm struggeling with the Kullback-Leibler divergence.