Re: [TLS] Require deterministic ECDSA

[Michael StJohns <[e-mail]>]

On 1/25/2016 7:41 PM, Bill Cox wrote:
> I have low expectations for IoT vendors' TRNGs.  When deadlines get 
> tight, good engineering on the TRNG is easy to drop.  As long as they 
> whiten the output, it is very difficult to detect TRNG flaws, so there 
> is little incentive to put in much engineering.  IIRC, the FIPS 
> standard requires vendors to be secretive about their TRNGs.  They are 
> not allowed to give us access to the raw data from the entropy source, 
> and cannot show us schematics for their design, making it nearly 
> impossible to differentiate a well designed TRNG from an insecure one.


Sorry for the late response on this one...

You should take a quick look at NIST Draft SP800-90B, section 7.1 
regarding how to do validation on entropy sources.    While this is 
still in draft form and doesn't yet trigger requirements in the FIPS 
validation process, I would expect it will at some point.   I would also 
expect that new designers are probably making sure that this type of 
interface is included in their products - to at least allow for the 
possibility of validation.

Of course, if an IoT vendor isn't looking for FIPS validation (or some 
other such process that requires at least a little testing), all bets 
are off.

Later, Mike