Re: [OAUTH-WG] RSA vs PKI?

Basically, the U.S. Government currently requires 256-bit symmetric
encryption keys for the highest levels of security (documents whose
confidentiality over the next several decades is a matter of national
security).

The 256-bit security requirement in symmetric keys compares to 512-bit in
ECC but only to something like RSA-15,000. It is quite clear that nobody
will ever deploy RSA-15,000 because it would be incredibly inefficient
compared to ECC-512. So there is no point in estipulating RSA key sizes for
the highest security settings.

Generally, n bits of symmetic key security map to 2n bits of ECC security
(currently published cryptanalytic research). However, the reduction of RSA
security to symmetric security does not follow a linear relationship.

On Sun, Nov 8, 2009 at 7:32 PM, Brian Eaton <beaton@google.com> wrote:

> On Sun, Nov 8, 2009 at 7:13 PM, Phillip Hallam-Baker <hallam@gmail.com>
> wrote:
> > Now RSA 2048 is almost certainly enough for most things, but when you
> > are a government agency you want the stuff you encrypt using RSA2048
> > on its last day of service life to be secure for another 20-30 years,
> > possibly more. So while RSA 2048 has at least another 30 years left in
> > it for authentication problems it is already reaching the end of its
> > service life for certain confidentiality applications.
> >
> >
> > I would suggest making RSA 2048 - SHA 256 the mandatory to implement
> > and chose an ECC alternative from suite B.
>
> So the choice of RSA key length seems like something that different
> applications could reasonably choose on their own, no?  Do we really
> need to include 2048 bit RSA in the spec, or is specifying RSA in
> general sufficient?
>
> I suspect we do need to specify acceptable hash output lengths if
> we're going to have decent interop.
>
> (I'm seriously pushing the bounds of my crypto knowledge here, so do
> let me know if any of the above is nonsense.)
>
> Cheers,
> Brian
> _______________________________________________
> OAuth mailing list
> OAuth@ietf.org
> https://www.ietf.org/mailman/listinfo/oauth
>

-- 
Breno de Medeiros