Re: [pkix] End-entity self-signed certs

["Miller, Timothy J." <[e-mail]>]

On 11/30/11 7:52 AM, "Tom Gindin" <[e-mail]> wrote:

>        The main point I've tried to make in this thread is the
>usefulness 
>of self-signed certificates for level 1 and as a credential after level 2
>proofing, rather than their unsuitability for higher assurance levels.
>Nonetheless, I don't know if anyone has ever used them at higher levels
>and I don't think they should.

To counter, I'm trying to make the point that functionally speaking the
self-signed certificate is nothing more than a container for the public
key.  A key continuity system wouldn't even look at the certificate
attributes.  Practically speaking, however, the certificate is needed
because the *protocols on which systems are built* are built around X.509
objects.

For example, I can (and have) build a web service that uses key continuity
for authN and authZ.  However, to *use* that service the user must have a
certificate, because the TLS protocol requires certificates to function.

>        The revocation requirements for levels 3 and 4 are IMHO a
>stronger 
>reason not to use self-signed certificates at those levels than the fact
>that a self-signed certificate consists mainly of information which
>nobody 
>except the claimant has vouched for.  The lack of revocation is
>irrelevant 
>to level 1, relevant but not crippling for level 2, serious for 3 and
>even 
>worse for 4.  You are right that a self-signed certificate could possibly
>represent a strong enough proofing process for level 3 or 4, although
>using one makes it hard for that fact to be conveyed between a CSP and a
>separate RP.

As you noted, revocation in key continuity systems are achieved by
de-authorizing the key.  I don¹t see this as a problem for LOA3/4, but it
does make the C&A job more important.

-- T