The attack vector is easier than that. 

1.      I use very stringent processes to verify that Google, Inc. is a legit 
company in Utah.
2.      I verify that Jeremy did indeed incorporate Google, Inc. 
3.      I call Jeremy at the phone listed for Google, Inc., the Utah corporation
4.      The domain information shows Google, Inc. as owning google.com
5.      Certificate issues.

 

Obviously this would be caught in every CA’s high risk checks, but the point 
remains valid. Regardless of the expertise and thoroughness of the org check, 
the specs lack any time between the verified org and the actual domain because 
orgs are not unique on a global basis.

 

Jeremy

 

From: Public [mailto:[email protected]] On Behalf Of Ryan Sleevi via 
Public
Sent: Thursday, December 21, 2017 9:58 AM
To: Adriano Santoni <[email protected]>; CA/Browser Forum Public 
Discussion List <[email protected]>
Subject: Re: [cabfpub] Verification of Domain Contact and Domain Authorization 
Document

 

Adriano,

 

Do you have an example of how you believe 3.2.2.4.1 can be used correctly? 

 

Specifically, it does not describe the process for validating that the 
Applicant is the Domain Contact with the Registrar - this isn't equivalent to 
using WHOIS.

 

Here's just one scenario:

- I ("Ryan Sleevi") apply to Foo CA for example.com <http://example.com> , 
which is owned by "Andriano Santoni's Lightly Validated Certificates" - you.

- Foo CA decides to employ 3.2.2.4.1, using 3.2.2.4(1)

  - Note, as worded, all of 3.2.2.1 can be read as 'optional' for DV certs, 
thus automatically met, but lets pretend its OV

  - They verify "Andriano Santoni's Lightly Validated Certificates" is a real 
company with a real existence using a QGIS. That's all that's needed - there's 
no binding to the Applicant, just an existence proof of the data.

  - Alternatively, I send a photoshopped letter claiming your company exists, 
valid under 3.2.2.1(4)

  - Alternatively, the CA declares that "Google Maps" is a Reliable Data Source 
(it isn't, but again, underspecified), and verifies that there's an entry under 
3.2.2.1(2) - despite the fact I just added the entry

- They then need to verify whether or not I'm authorized to speak for your 
company.

  - The information used in 3.2.2.1 doesn't have to be used ("the CA MAY use 
..."), but remember, I may have made it up under 3.2.2.1

  - The CA can directly call me, Ryan Sleevi, asking if I'm authorized ("the CA 
MAY establish the authenticity of the certificate request directly with the 
Applicant Representative")

  - The requirement to use an RMOC simply means that Foo CA could decide to 
call up Jeremy, since Jeremy knows me, and say "Hey, does Ryan work for Adriano 
Santoni" - that's all that's required.

- Finally, the CA contacts the registrar, and says "Hey, does Adriano Santoni's 
Lightly Validated Certificates own example.com <http://example.com> " - and the 
registrar says sure

  - Note: There's no consensus whether we're talking about the same 
organization - perhaps I created a version incorporated in the US, but you're 
incorporated in Italy

 

These are just a few of the legal-but-bad things you can do. I'm sure we'll see 
the normal rush from some CAs saying "Yes, but we'd never do that" - while 
ignoring the fact that some could, as it's valid under the language, and we 
consistently see "That which is valid (or subject to misinterpretation) is 
possible to use"

 

 

Could you provide an example of how you believe 3.2.2.4.1 "should" work and 
offer the same level of assurance as the other methods, without normatively 
prescribing the data sources used? From conversations with both current and 
past employees of CAs, I am adamantly convinced that there is not a consistent 
standard of reliableness being applies. Google Maps being used as a Reliable 
Data Source is not a hypothetical, despite it allowing community edits.

 

 

On Thu, Dec 21, 2017 at 4:00 AM, Adriano Santoni via Public 
<[email protected] <mailto:[email protected]> > wrote:

Jeremy, I am not sure I fully understand the problems you describe. 

Would it be possible for you to provide some concrete example related to method 
#1, with some details, without of course mentioning specific certificates 
and/or organizations?

 

Il 19/12/2017 22:30, Jeremy Rowley via Public ha scritto:

Hi all, 

 

When reviewing the Symantec validation methods and the customers using each 
method, I found an alarming number of customers verified under 3.2.2.4.1 
(Verification of a Domain Contact) or 3.2.2.4.5 (Domain Authorization Document) 
where the domain is not technically associated with the entity. These two 
methods need improvement or removal as the way they are currently lacks 
sufficient controls to associate the domain verification with the actual 
certificate approver. I’ve had too many calls with customers explaining 
re-verification where the domain holder didn’t understand that a cert issued 
for the domain. Although the organization verification was successfully 
complete, the only tie between the domain and organization is a call to the 
organization that happened within the last years to approve the account for 
issuance. I wanted to bring it up here because I’ve always thought these 
methods were less desirable than others. I think other large CAs use this 
method quite a bit so I’m hoping to get clarity on why these methods are 
permitted when the domain verification seems more “hand-wavy” than other 
methods. 

 

Method 3.2.2.4.1 permits a CA to issue a certificate if the certificate is an 
EV or OV cert. With EV certificates, there is a call to a verified telephone 
number that confirms the requester’s affiliation with the organization. I can 
see this method working for EV.  For OV certificates, there is a reliable 
method of communication that confirms the account holder as affiliated with the 
organization.  Unlike EV, for OV certs there is no tie between the requester 
and their authority to request a certificate. Once the organization is 
verified, the BRs permit auto-issuance for any domain that reflects an 
affiliation with the verified entity for up to 825 days. There’s no notice to 
the domain contact that the certificate was requested or approved.  Perhaps 
this is sufficient as the account has been affiliated with the organization 
through the reliable method of communication and because CT will soon become 
mandatory. 

 

Method 3.2.2.4.5 permits a CA to issue a certificate using a legal opinion 
letter for the domain. Unfortunately the BRs lack clear requirements about how 
the legal opinion letter is verified. If I want a cert for Google.com and the 
CA is following the bare minimum, all I need to do is copy their letterhead and 
sign the document. Magically, a certificate can issue.  This method lacks a lot 
of controls of method 1 because there is no requirement around verification of 
the company. I can list as many domains in the letter as I’d like provided the 
entity listed in the corresponding WHOIS’s letterhead is used.

 

I’m looking to remove/fix both of these methods as both these methods lack the 
necessary controls to ensure that the verification ties to the domain holder. 
These methods probably should have been removed back when we passed 169/182. 
Would anyone being willing to endorse a ballot killing these or making some 
necessary improvements?  

 

Jeremy

 

 

_______________________________________________
Public mailing list
[email protected] <mailto:[email protected]> 
https://cabforum.org/mailman/listinfo/public

 


_______________________________________________
Public mailing list
[email protected] <mailto:[email protected]> 
https://cabforum.org/mailman/listinfo/public

 

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Public mailing list
[email protected]
https://cabforum.org/mailman/listinfo/public

Reply via email to