No-leak authentication by the Sherlock Holmes method

Abstract. We propose a class of authentication schemes that are literally zero-knowledge, as compared to what is formally defined as “zero-knowledge” in cryptographic literature. We call this “no-leak” authentication to distinguish from an established “zero-knowledge” concept. The “no-leak” condition implies “zero-knowledge” (even “perfect zero-knowledge”), but it is actually stronger, as we illustrate by examples. The principal idea behind our schemes is: the verifier challenges the prover with questions that he (the verifier) already knows answers to; therefore, even a computationally unbounded verifier who follows the protocol cannot possibly learn anything new during any number of authentication sessions. This is therefore also true for a computationally unbounded passive adversary.