[Cryptech Tech] RNG test tools wiki page

[Bernd Paysan]

Am Freitag, 1. August 2014, 22:41:53 schrieb Benedikt Stockebrand:
> > There are tests in dieharder which are sensitive to very similar things,
> > and will have corellated results.  Putting these tests together into a
> > too tough meta-test wouldn't be a good idea. Putting independent tests
> > (tests which test different properties of randomness) into a meta-test
> > IMHO is good.
> I can't quite follow here: If one test shows that there is a significant
> bit bias, why would I combine it with one testing for correlation with a
> bit distance of 8, and therefore dilute the result?

No, you perform a meta-test, that is you test how a larger number of tests 
work.  The idea here is that all these tests test independent statistical 
properties, so they should behave like independent variables, and such 
independent variables in a random sample have properties you can analyze.  
E.g. they distribute evenly.  And they change independent when used on new, 
independent data.

> > You'll also expect that the same tests run on an independent data sample
> > will have independent results.
> 
> Sorry, maybe I'm just too tired (and still struggling with a cold from
> that polar airport air...), but what exactly do you mean with that?

Ok, you collect e.g. 1MB data from a random number generator.  And then you 
take another 1MB data from that random number generator, and another and so on 
(all those are supposed to be independent data samples).  Now you can perform 
the same tests on both data, and record the results.  There will be a result 
"Bias is 50%+-some small number" and a result "correlation of m bits over n 
bits period distributed into n buckets is such and such".  If you look at 
those results for all those independent 1MB data samples you have taken, they 
should distribute according to laws of statistics.

-- 
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://bernd-paysan.de/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.cryptech.is/archives/tech/attachments/20140802/35822598/attachment.sig>