Re: [pkix] RFC 5280 interpretation of trust anchor certificates

[Jeffrey Walton <[e-mail]>]

On Sun, Oct 9, 2022 at 1:53 PM Niklas Matthies <[e-mail]> wrote:
>
> Dear all,
>
> On the ESI (ETSI) mailing list, the question came up whether RFC 5280
> says anything about trust anchors provided in the form of certificates
> that are _not_ self-signed.
>
> In section 6.1.1, there is the following wording on page 76:
>
>        The trust anchor information may be provided to the path
>        processing procedure in the form of a self-signed certificate.
>        When the trust anchor information is provided in the form of a
>        certificate, the name in the subject field is used as the trusted
>        issuer name and the contents of the subjectPublicKeyInfo field is
>        used as the source of the trusted public key algorithm and the
>        trusted public key.
>
> The first sentence seems to indicate that the case of providing a
> trust anchor in the form of a non-self-signed certicate is not
> considered here. But the second sentence doesn't repeat the
> "self-signed" bit, which can be interpreted as that sentence also
> applying to non-self-signed certificates. However, if that is the
> case, why does the first sentence restrict itself to specifically
> self-signed certificates?
>
> On page 74 there is also the following wording:
>
>        When the trust anchor is provided in the form of a self-signed
>        certificate, this self-signed certificate is not included as
>        part of the prospective certification path.
>
> If RFC 5280 also allows the possibility of trust anchors being
> provided in the form of non-self-signed certificates, then it would
> seem that the above restriction would not apply to those, i.e., that
> they may be included as part of the prospective certifcation path.
> However, I don't see how that would make any sense.
>
> All the wording taken together, my conclusion up to now was that RFC
> 5280 simply does not consider the possibility that the trust anchor
> could be provided in the form of a non-self-signed certificate, and
> that therefore, specifications which *do* allow for that possibility
> (such as in the context of ETSI trusted lists) have to clarify how
> that case maps onto what RFC 5280 expects.
>
> If that interpretation is incorrect, that is, if RFC 5280 doesn't
> actually care about whether a trust-anchor-representing certificate
> provided as input to the path validation algorithm is self-signed or
> not, then maybe an erratum is in order?
>

Also see the path building RFC - RFC 4518. Unfortunately, the path
building RFC uses self-signed CA certificates or Root CA certificates
for trust anchors.

I generally ignore that requirement since it increases attack surface
and risk. I prefer to root trust in a CA certificate that is closest
to the end-entity certificate. It prunes the tree and attack surface
and reduces risk.

In the past, under RFC 14158, you could not anchor trust in Let's
Encrypt. You had to use one of the CA's that certified Let's Encrypt.
I don't know if it's still the case.

RFC 4518, Internet X.509 Public Key Infrastructure: Certification Path
Building, https://www.rfc-editor.org/rfc/rfc4158 .

Jeff