h, then there is a security connection between X and Y. For a cryptographically-secure pseudo-random number generator (Manuel Blum and Silvio Micali analyzed complexity-theoretic aspects of this, relative to the above paper), this should be fine, even though I usually suggest independent keys.