Is it possible to directly compute the set-difference of two pointset powersets?

In particular, I’m interested in the precise complement; 

convex difference approximations are not sufficient for my case.

Kind regards,

Jaco van de Pol

--