Statistically, any routine which results in randomized output from non-random input (and especially if it is repeatably deterministic) would be a good candidate.