[Cryptech Core] capacity/performance numbers?

[Phil Roberts]

Thanks for the clarifications.  I think the immediate request was for
support of up to 2048.


On 6/24/16, 12:46 PM, "Rob Austein" <sra at hactrn.net> wrote:

>At Fri, 24 Jun 2016 17:32:44 +0000, Phil Roberts wrote:
>> 
>> Thanks for this.  A related question came up in terms of what RSA key
>> lengths we support.  Should I interpret this to mean that we support
>>both
>> 4096 bit and 8192 bit keys today?
>
>Yes and no.
>
>The code is designed to support RSA keys in the range 1024-8192 bits.
>
>The math library we're using requires a compile-time declaration of
>the largest bignum one intends to use (price tag for avoiding all use
>of dynamic memory in the math library, seems like a fair trade), and
>we have not tested anything above 4096 bits.
>
>But the real limiting factor is probably speed.  Certainly key
>generation for anything above 2048 bits is painfully slow at the
>moment.  Most of our tests involve both key generation and signature,
>and I haven't done enough signature tests with pre-generated keys to
>have a good feel for how fast signing is without key generation.