[Cryptography] TRNG review: Arduino based TRNGs

[Bill Cox]

On Fri, Jan 15, 2016 at 9:05 AM, Ron Garret <ron at flownet.com> wrote:

>
>
> Even in that case why not give yourself a comfortable safety margin?  Are
> there really applications that require you to squeeze every last bit of
> entropy out of a noise source?
>
> rg
>
>
It's not the rate of entropy collection, but whether there is _any_ entropy
coming from the source.  A 256 bit key is going to be pretty weak if is a
single 8-bit constant repeated 32 times.  If the input on A0 is
consistently in the center of a 2mV range that is output as the same
constant on each read, that's what you'll get.

Even with a simple circuit and PWM feedback to ensure that the noise
voltage will cause the input voltage to cross boundaries between 2mV
ranges, and be recorded as one of the values randomly, this circuit remains
highly sensitive, in that any injected signal on the order of the noise
amplitude (24uV in my example) can control the output, overriding any real
randomness.  We can design around this constraint, with reasonable
shielding, supply regulation, bypass caps, and such, as appropriate for the
application.  It just takes good engineering.

Too bad this $0.001 worth of hardware isn't pre-engineered for us ;)

Bill
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20160115/5f73d222/attachment.html>