[cabfpub] Ballot 218: Remove validation methods #1 and #5

Wed Jan 10 23:49:20 UTC 2018

So to make sure I've captured the discussion points of 3.2.2.4.1 for the
"things that would be disruptive"

For situations like GoDaddy (in which the CA is the Registrar as well) - or
for situations like, say, Google Trust Services/Google, in which the CA is
an Affiliate of the Registrar (I think; I'm sure folks like Amazon with its
Amazon Trust Services is in a similar space, and no doubt others) - the
'goal' is that if the CA believes the Applicant is the same entity as the
Domain Owner, then it can use that as an effective authorization by virtue
of the fact that it's the same account that would approve that operation.

For situations like the ccTLDs, particularly those that lack WHOIS, or for
situations like .gov, it's seen as acceptable that contacting the registry
to determine who the Domain Name Registrant is, and then using
3.2.2.4.2/3.2.2.4.3 methods of validation/contact is seen as viable.

Finally, we have the issue that most folks (all?) seem to agree that
3.2.2.4.1 Option 1/Option 2 are not reliably secure, and neither is
3.2.2.5. Unfortunately, because of the 'grandfathering' validation clause,
this makes it difficult to unambiguously ensure (re)validation of domains
if they use an insecure method.

Daymion, do you think a ballot that:
- Deprecated 3.2.2.4.1 as previously proposed (e.g. "adding not after
March")
- Deprecated 3.2.2.4.5 as previously proposed (e.g. "adding not after
March")
- Make the change to Contact as previously discussed, namely: "Domain
Contact: The Domain Name Registrant, technical contact, or administrative
contract (or the equivalent under a ccTLD) as listed in the WHOIS record of
the Base Domain Name or in a DNS SOA record, or as obtained through direct
contact with the Domain Name Registrar."
- Introduce 3.2.2.4.11 (a new method), which reads (to be smithed a little
if necessary):

"3.2.2.4.11 Validating Applicant as a Domain Contact
Confirming the Applicant's control over the FQDN by validating the
Applicant is the Domain Contact. This method may only be used if the CA is
also the Domain Name Registrar, or an Affiliate of the Registrar, of the
Base Domain Name.

Note: Once the FQDN has been validated using this method, the CA MAY also
issue Certificates for other FQDNs that end with all the labels of the
validated FQDN. This method is suitable for validating Wildcard Domain
Names."

Yes, this doesn't describe how the CA/Registrar (or CA-Affiliate/Registrar)
precisely performs this validation, but I'm hesitant to add wording such as
'accounts' or 'usernames' or equivalent access control levels at this time,
and can live with being permissive here and then iterating to be tighter
around this language. My goal here would be to make it clear that Option #1
and Option #2 of 3.2.2.4.1 are unacceptable, but unfortunately the poor
wording of 3.2.2.4 and 4.2.1 means we have to effectively remove/forbid
3.2.2.4.1 entirely and then re-add the bits that make sense.

On Wed, Jan 10, 2018 at 6:16 PM, Daymion T. Reynolds <dreynolds at godaddy.com>
wrote:

> Ryan,
>
>
>
> Q: Can you explain why you do not believe it is more secure?
>
> A: I am not stating its more secure, but .1 Option #3 is on par with other
> deemed secure options. It is secure because only the domain owner is the
> authorized user to a registrars account, and only they can order a cert for
> a specific domain. If more transparency is a concern, let’s discuss what
> would be acceptable.
>
>
>
> Q: I don't believe this is universally the case. Consider the situation of
> registrars and registries that allow signing in without a 2FA, but require
> changes to use a 2FA. I realize a response might be "Well, the registrar
> could just require 2FA for issuing a cert" - and while that would be in
> theory possible, there's absolutely zero assurance for relying parties and
> browsers as to the registrars (and CAs) practices. I hope you can see why
> this remains a fundamentally problematic proposal.
>
> A: I agree, everyone should be using 2FA. Unpacking this a bit further, as
> your statement is close to the crux of my argument. If the registrar
> account is compromised, everything about the domain is in question. If
> registrar account compromise is a concern for .1 option #3 then it also a
> question for .4 Domain Contact, .7 DNS Change and .8 IP Address validation.
> All of these require the registrar account to be secure. I believe .4, .7,
> and .8 are secure practices.
>
>
>
> Q: I think it's important to be precise here when talking about .1. Is it
> correct to say you are only concerned with retaining some notion of .1
> Option 3? When we say "don't eliminate .1", I think that carries with it
> the significant (and insecure) suggestion of retaining .1 Option 1 and .1
> Option 2.
>
> A: Agreed, and sorry for not being precise. You are correct I am only
> concerned with Option #3, as I believe it is a secure practice on par with
> .7 and .8. I agree with your position on option #1 & #2.
>
>
>
> Thanks for your time discussing this,
>
> Daymion
>
>
>
>
>
>
>
>
>
> *From:* Ryan Sleevi [mailto:sleevi at google.com]
> *Sent:* Wednesday, January 10, 2018 1:50 PM
> *To:* Daymion T. Reynolds <dreynolds at godaddy.com>
> *Cc:* CA/Browser Forum Public Discussion List <public at cabforum.org>; Kirk
> Hall <Kirk.Hall at entrustdatacard.com>; Tim Hollebeek <
> tim.hollebeek at digicert.com>
> *Subject:* Re: [cabfpub] Ballot 218: Remove validation methods #1 and #5
>
>
>
>
>
>
>
> On Wed, Jan 10, 2018 at 2:37 PM, Daymion T. Reynolds <
> dreynolds at godaddy.com> wrote:
>
> Ryan,
>
>               Thank you for replying as this is a good discussion to have.
> “Direct contact” is great method when you don’t have a clean, reliable data
> source to validate ownership. For Registrar / CA combos, whereby the same
> account ordered the domain and the cert, knowledge of ownership is robust.
> Requiring a second contact doesn’t seem more secure, but rather seems more
> cumbersome for an already complex process.
>
>
>
> Can you explain why you do not believe it is more secure?
>
>
>
> If you are concerned about the possibility of a customer account being
> compromised, it doesn’t change the risk. If there was a compromise they
> would have control over DNS and could then domain validate a cert order
> from anyone.
>
>
>
> I don't believe this is universally the case. Consider the situation of
> registrars and registries that allow signing in without a 2FA, but require
> changes to use a 2FA. I realize a response might be "Well, the registrar
> could just require 2FA for issuing a cert" - and while that would be in
> theory possible, there's absolutely zero assurance for relying parties and
> browsers as to the registrars (and CAs) practices. I hope you can see why
> this remains a fundamentally problematic proposal.
>
>
>
>               Rather than eliminate .1, I believe a better course of
> action would be to add transparency and lock down when you can and cannot
> use the registrar validation method.
>
>
>
> I think it's important to be precise here when talking about .1. Is it
> correct to say you are only concerned with retaining some notion of .1
> Option 3? When we say "don't eliminate .1", I think that carries with it
> the significant (and insecure) suggestion of retaining .1 Option 1 and .1
> Option 2.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cabforum.org/pipermail/public/attachments/20180110/86271daa/attachment-0003.html>