[Cryptech Tech] Dieharder test of Cryptech RNG data

[Benedikt Stockebrand]

Hi folks,

Joachim Strömbergson <joachim at secworks.se> writes:

> Here is the test report from running the (-a)ll test suite in Dieharder
> on abou 18 GBytes of RNG data extracted yesterday. I will extract a
> bigger dataset and do new tests as suggeted by Benedict later. But imho
> this looks fairly good. No test case failed, only one classed as weak.

you should expect one to be weak on a regular run; it is most convenient
to use "-Y 1" ("resolve ambiguity mode") and feed it as much data as
needed to get a more decisive result; and don't forget to add a "-k 2"
as per the man page as well.  Problem then of course is that it may run
out of test data...

However, the more important thing to keep in mind is that dieharder is
aimed at general-purpose (i.e. non-crypto) algorithmic pseudo random
number generators, so running it against a TRNG will only show rather
blatant mistakes.

> (Dieharder really benefits from fast CPU and lots of mem. Running this
> on my laptop takes hours. On the quad core monster w 32 GBytes ram in
> the basement this took about 20 minutes. ;-)

That's good news.  I somewhat suspected so from the way it was written;
my guess is that it's still running on a single CPU thread only, but at
least the RAM keeps everything in I/O cache rather than reading it from
disk for every single test.


BTW, there's sort of a chance that I'll find some time to work on my
test framework this month.  I'll keep you posted if/when anything like
some sort of progress happens.


Cheers,

    Benedikt

-- 
Benedikt Stockebrand,                   Stepladder IT Training+Consulting
Dipl.-Inform.                           http://www.stepladder-it.com/

          Business Grade IPv6 --- Consulting, Training, Projects

BIVBlog---Benedikt's IT Video Blog: http://www.stepladder-it.com/bivblog/