[Cryptech Tech] Maurer's Universal Test for Randomness

[Peter Gutmann]

=?UTF-8?B?Sm9hY2hpbSBTdHLDtm1iZXJnc29u?= <joachim at secworks.se> writes:

>While reading a protocol spec at work I found a reference to the following
>paper. The paper describes a universal test for randomness

That doesn't mean what you think it means.  A universal probability
distribution means that if you compare it to another distribution, it's not
lower by any more than some constant factor, however the constant depends
entirely on what two distributions are being compared.  So a universal entropy
estimator overestimates the entropy by no more than a constant, but the
constant is unbounded.

Peter.