Re: [dsfjdssdfsd] specifying an RNG

[Donald Eastlake <[e-mail]>]

Hi,

On Fri, Nov 15, 2013 at 5:55 AM, Eric Burger <[e-mail]>
wrote:
> On Nov 15, 2013, at 12:55 AM, Dan Harkins <[e-mail]> wrote:
>>  Hello, and welcome to the dsfjdssdfsd list!
>>
>>  At the last IETF the question was asked, "what can we do to harden
>> the Internet?" Given the recent news about Dual_EC_DBRG and the
>> dopant attack against hardware RNGs one of the things that can be
>> done is to have an open specification of a secure RNG.
>
> I like this idea.
>
>> This would
>> allow developers to have an alternative to relying solely on
>> /dev/[u]random, a hardware RNG, an RNG specified by a large
>> government-affiliated group to mix the uncorrelated sources of
>> entropy they gather, or an RNG designed in an ad hoc manner by
>> someone who thinks he knows what he's doing but probably
>> doesn’t.
>
> History is nice, but if we produce a *standard*, one would hope the
developer goes to /dev/IETFrandom. I.e., none of the issues listed above
are relevant.
>
>>  One of the things that would be nice to get out of this list is a
>> specification on a strong RNG, in the form of a BCP or Informational
>> RFC. This doesn't necessarily mean lets "roll our own" but perhaps
>> there is best practice that can be specified.

RFC 4086, which is on this topic, is a BCP but is out of date and doesn't
recommend a specific random number generator.
https://datatracker.ietf.org/doc/draft-eastlake-randomness3/
is the start of an effort towards a replacement but needs a lot of work.

By the way, not that I am arguing for or against it for this case, but it
is quite possible for a BCP or IETF standard to consist of a set of RFCs…

Thanks,
Donald
=============================
 Donald E. Eastlake 3rd   [phone] (cell)
 155 Beaver Street, Milford, MA 01757 USA
 [e-mail]

> How is this any different than today’s practice? Today we point to NIST,
FIPS, NSA, RSA, ISO, Government Committee of the Russia for Standards, etc.
>
>>  So, is there a model that defines what a "robust RNG" would look
>> like? Is there a suitable candidate that exists already for such a thing?
>
> I am still having trouble understanding what would be different than what
we do today. If you are proposing something akin to the open source codec
work, then I am only just a bit skeptical, but I was proven wrong by that
effort (it seems to be succeeding). However, if all we are doing is looking
for a RNG to put into IETF standards, then are we not doing that already?