Correlated Product Security From Any One-Way Function and the New Notion of Decisional Correlated Product Security

It is well-known that the k-wise product of one-way functions remains one-way, but may no longer be when the k inputs are correlated. At TCC 2009, Rosen and Segev introduced a new notion known as Correlated Product secure functions. These functions have the property that a k-wise product of them remains one-way even under correlated inputs. Rosen and Segev gave a construction of injective trapdoor functions which were correlated product secure from the existence of Lossy Trapdoor Functions (introduced by Peikert and Waters in STOC 2008). The rst main result of this work shows the surprising fact that a family of correlated product secure functions can be constructed from any one-way function. Because correlated product secure functions are trivially one-way, this shows an equivalence between the existence of these two cryptographic primitives. In the second main result of this work, we consider a natural decisional variant of correlated product security. Roughly, a family of functions are Decisional Correlated Product (DCP) secure if f1(x1);:::;fk(x1) is indistinguishable from f1(x1);:::;fk(xk) when x1;:::;xk are chosen uniformly at random. We argue that the notion of Decisional Correlated Product security is a very natural one. To this end, we show a parallel from the Discrete Log Problem and Decision Die-Hellman Problem to Correlated Product security and its decisional variant. This intuition gives very simple constructions of PRGs and IND-CPA encryption from DCP secure functions. Furthermore, we strengthen our rst result by showing that the existence of DCP secure one-way functions is also equivalent to the