Two long-standing open problems exist on the fringe of Complexity Theory and Cryptography: 1. Does there exist a reduction from an NP-Complete Problem to a one-way function? 2. Do parallelized versions of classical constant-round zero-knowledge proofs for NP conceal every "hard" bit of the witness to the statement proved? We show that, unless the Polynomial-Hierarchy collapses, black-box reductions cannot be used to provide positive answers to both questions.