[Cryptech Tech] Noise board on Novena

[Fredrik Thulin]

On Saturday, September 20, 2014 11:18:00 AM Benedikt Stockebrand wrote:
...
> > We haven't gotten to sort of specifying what quality the FPGA will
> > expect/require from the external entropy source(s). I'd say it probably
> > won't be "must be perfectly random", but I don't know how we will define
> > it.
> > 
> > As food for thought, maybe the spec would be something like "must give a
> > Monte Carlo test value of less than 0.1% in 100k readings" (I totally
> > just made up those numbers).
> 
> As I mentioned before, to my understanding that makes testing/auditing
> significantly more difficult.  But yes, we definitely have to work on
> that some time.

To clarify, I think such a requirement would be a written requirement directed 
to the designers of 3rd party entropy sources, not an FPGA enforced one. It is 
my understanding too that it is rather expensive to implement statistical 
tests over long series of output from the entropy sources in the FPGA.

I think the always-running tests of the entropy sources will be limited to 
making sure they are not stuck at some value, and maybe a test to make sure 
the overall distribution of 0 and 1 bits are somewhat close to 50%.

/Fredrik