If nobody can prove (even in an all-against-one cooperation) that one did not vote with a particular cast, then one can claim
anything about his cast even under oath, and has no fear of being caught. We consider the question of constructing a voting
scheme that provides all participants with this “absolute” privacy.
We assume that half of the problem is already solved: The votes are evaluated so that only the result is revealed. Latest
achievements of secure coprocessors are supposedly a justification for such a presumption. We prove that even under the presumption
that the voting reveals nothing but a result, the privacy of an individual input can withstand an “all-against-one” attack
under certain conditions only.
First condition: The function that maps a set of casts to the result of voting must be non-deterministic. Second condition
(paradoxically): for any set of casts any result must be possible.