Re: [Cfrg] FIPS or equivalent approvals

[David Jacobson <[e-mail]>]

On 7/30/14 8:54 PM, David Jacobson wrote:
> On 7/29/14 11:43 AM, Salz, Rich wrote:
>>> I feel a HSM/smartcard/token design actually worthy of
>>> third-party trust would need:
>> Check out http://cryptech.is
>>
>> -- 
>> Principal Security Engineer
>> Akamai Technologies, Cambridge MA
>> IM: [e-mail] Twitter: RichSalz
>>
> In a past life, I worked on the design of an HSM that was used by lots 
> of banks.  (FIPS 140-2 level 3)   It was basically just an ARM 
> processor and a crypto accelerator in a module coated with hard opaque 
> epoxy.  There is no reason that the firmware couldn't be modified to 
> do Curve 25519.  Of course, it would be slower, since it wouldn't get 
> any benefit from the crypto accelerator, and you couldn't put it in 
> FIPS mode with that curve enabled (not a "FIPS approved algorithm").
>
>     --David Jacobson
Oops.  I replied to the wrong message.  There was a talk of how a new 
curve would require new HSMs, and the manufacturers should get working 
on it, etc.  My point was that for the manufacturer who has all the code 
for the HSM, adding a new curve is is much simpler than designing an HSM 
from scratch.   Rich's remark was about the need for an open HSM, so we 
could trust it.  An entirely different subject.

    --David