[Cryptech Tech] Dieharder results for Cryptech RNG data

[Benedikt Stockebrand]

Hi Joachim and list,

Joachim Strömbergson <joachim at secworks.se> writes:

> Sorry for a late reply.

no need to apologize to *me* on that:-/

>> - What options did you run dieharder with exactly?
>
> -a -g 201
>
> That is all default test cases and external binary data.

OK, for those settings it needs about 13+ GB or so of data.  The problem
here is that some tests trigger rather sensitively for reused data,
while others don't.  In the end, the only way to make any reliable use
of dieharder is to feed it all the data it needs.

>> - What was the test set size? 200+ MB?  That's a long way short of 
>> what you need for reliable results; if you did the full run, that 
>> rgb_lagged_sum with ntuples==15 looks really bad.
>
> That was the amount of data I had extracted at the time. I have more but
> need to change the HW design to get better throughput. My blinkenlights
> is hogging the output.

I know that problem:-) When I realized that my big test series failed
because I had used the wrong MCU firmware, that was more than three
weeks worth of test data.  No relation to the name of the "ARRGH"
board...

But anyway, yes, we actually need decent throughput if alone for running
these tests.  And I think we also need a test suite targeted at HWRNGs,
so we can get usable test results with less test data.  Dieharder was
written for PRNGs and the assumption that these are pretty much
arbitrarily fast, so while it is sure useful as a "not written here"
test suite, it isn't really all that well suited for testing during
development.


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/