[Cryptography] Best internet crypto clock

dj at deadhat.com dj at deadhat.com
Thu Oct 23 13:43:49 EDT 2014


> On Wed, Oct 22, 2014 at 10:35 AM, Henry Baker <hbaker1 at pipeline.com>
> wrote:
>
>> Perhaps a better source would be something that couldn't possibly be
>> hacked -- e.g., variations in solar flux of neutrinos or other solar
>> variations.
>
> How about using the earth's rotation? :)
>
> http://www.iers.org/IERS/EN/DataProducts/EarthOrientationData/eop.html;jsessionid=5F896E89B594D8D5B7A9E38DE4AD6BF0.live1
>
> Only Superman can hack that...
> _______________________________________________

Remote sources require external inputs.

Transistors have plenty of thermal noise in their gates. It's local, well
understood and can be modeled for min-entropy analysis over all
environmental conditions and attack scenarios.

You just need know how to get at it in a robust way. We published two
different circuits that do this. The one for RdRand that's been repeated
in numerous papers (like this:
http://www.cryptography.com/public/pdf/Intel_TRNG_Report_20120312.pdf) and
the other is this one:
http://www.researchgate.net/publication/224170854_2.4GHz_7mW_all-digital_PVT-variation_tolerant_True_Random_Number_Generator_in_45nm_CMOS

There are many gigabits/s of data you can get out of a transistor with a
high entropy distribution. You can be reasonably confident that the noise
in the transistor gate is the aggregate signal from many quantum events in
the particles out of which it is constructed.

On silicon, the circuits are small. You may be able to do something
similar with discrete components, albeit at lower speed.




More information about the cryptography mailing list