[cabfpub] Concerns regarding Mozilla Root Program/Baseline Requirements

Jeremy Rowley jeremy.rowley at digicert.com
Fri Aug 2 23:01:48 UTC 2013


Agreed - with the unknown definition of rekey, you can have a completely
difference cert issued than the previous one.  

Plus, permitting this type of behavior becomes a nightmare for detection. If
a CA can freely "re issue" a non-compliant cert with a notBefore date that
is after the BRs, no one looking at the industry can tell when a CA is
mis-issuing versus simply rekeying an old legacy cert.  By permitting this,
the CAB Forum would effectively nullify all monitoring of BR compliance as
any CA.  Any CA caught not complying with the BRs can simply claim a
suspicious cert as reissuance of an existing cert to escape severe public
scrutiny, and may even do so ignorantly.  Therefore, from a practical
standpoint, anything issued after the BR effective date really must comply
with the BRs, regardless of whether it’s a rekey, renewal, or modification.


Jeremy

-----Original Message-----
From: Ryan Sleevi [mailto:sleevi at google.com] 
Sent: Friday, August 02, 2013 3:56 PM
To: kirk_hall at trendmicro.com
Cc: Steve Roylance; Rich Smith; jeremy.rowley at digicert.com; Sigbjørn Vik;
public at cabforum.org
Subject: Re: [cabfpub] Concerns regarding Mozilla Root Program/Baseline
Requirements

On Fri, Aug 2, 2013 at 2:28 PM, kirk_hall at trendmicro.com
<kirk_hall at trendmicro.com> wrote:
> We also agree.  We were part of all BR discussions, and the effect of
rekeying was never discussed.  If it had been, our position would have been
that rekeying would not make a cert that was issued before the BRs become
subject to the BRs -- that would put CAs in the position of having to breach
their valid pre-BR contracts (which promised rekeying if needed) with their
customers -- something no CA would have agreed to, barring a showing of an
extraordinary security risk which we don't have here.
>
> Disclosure:  we have no certificates affected by this discussion, but
think the Forum would do better to focus on actual, current risks that need
to be remedied, and let the improvements we have already made through the
BRs take effect as older certificates expire by their terms.

Kirk,

I think the clearly fundamental disagreement on 'issuance' puts at risk
every other advance we might make in the BRs to address the actual, current
risks, which is why we raised this issue and are thankful to see the
spirited discussion.

It's been made abundantly clear that, during discussions on expirations, CAs
were implicitly assuming that re-keying and re-issuance would be exempt from
the BRs. However, I think that highlights the challenge - if that's a valid
interpretation, there's absolutely nothing in the BRs to suggest it's "dates
only" - after all, why not issue arbitrary extensions, unverified names,
etc? Why not add new fields or extensions? After all, if the pre-BR
contracts promised certain functionality that the BRs prohibit - whether it
be expiration or any other aspect - why wouldn't this be acceptable?

You can see that, at a minimum, we should least converge on an acceptable
definition about what certificate fields can be manipulated, and what the CA
is obligated to do when issuing new certificates under pre-BR contracts.

While not exactly a "Yes" vote for Rich's proposal, I think it's certainly
the most reasonable attempt to close this hole.

I just hope and expect CAs are fully aware and considering the implications
of the language - as it appears they may not have when negotiating the BRs -
and how this might affect other provisions:
domain validation, subject information, EKUs, revocation responders, serial
numbers/entropy, etc. It's entirely possible that such 'reissued'
certificates will differ much more significantly than just flipping some
dates and/or a key.

Cheers,
Ryan

>
> -----Original Message-----
> From: public-bounces at cabforum.org [mailto:public-bounces at cabforum.org] 
> On Behalf Of Steve Roylance
> Sent: Friday, August 02, 2013 6:46 AM
> To: Rich Smith; jeremy.rowley at digicert.com; 'Sigbjørn Vik'; 
> public at cabforum.org
> Subject: Re: [cabfpub] Concerns regarding Mozilla Root 
> Program/Baseline Requirements
>
> +1
>
> ..and just to be clear, we are talking about a re-key where only the key
(and therefore SKI) change as well as the unique elements of serial number
> and start date.   All other information remains the same as the previous
> certificate.    This is what the majority of players in the CA industry
> today deem "re-issuance" and it's mainly due to the complexity of PKI on
platforms coupled with the increase in virtualisation where keys seem to be
lost more frequently that has caused the CA industry to offer these types of
services to help PKI grow - without the need for customers to buy separate
insurance cover to help pay for this service or be charged a 2nd
> time for accidental loss.   It's not something one CA can take away by
> itself and it's not easily something one CA can change by itself, but it's
one area we all seemed to miss when looking at reduction of maximum
certificate duration from 60 to 39 so that should be the focus (As Rich
states and as I identified).
>
>
> Steve
>
>
> On 02/08/2013 14:30, "Rich Smith" <richard.smith at comodo.com> wrote:
>
>>Jeremy,
>>I agree with your reasoning, but at the same time, I don't think that 
>>we ever really thought in all the way through regarding the 
>>re-key/reissuance of certificates which had already been sold.  As I 
>>said in my recent reply to Sigbjorn, I'm not really all that concerned 
>>with certs sold for longer than 60 month term prior to the BRs.  If 
>>the consensus is that those need to have time cut off, I can live with 
>>that, but I would like to see us come up with a reasonable compromise 
>>to deal with those certs which are fully under the BR umbrella that 
>>fall into the 60 to 39 month change over period.
>>-Rich
>>
>>> -----Original Message-----
>>> From: public-bounces at cabforum.org
>>> [mailto:public-bounces at cabforum.org]
>>> On Behalf Of Jeremy Rowley
>>> Sent: Friday, August 02, 2013 7:26 AM
>>> To: 'Sigbjørn Vik'; public at cabforum.org
>>> Subject: Re: [cabfpub] Concerns regarding Mozilla Root 
>>> Program/Baseline Requirements
>>>
>>> As pointed out, a ten year or five year timeline for implementing 
>>> change in the industry is very long.  This was discussed extensively 
>>> during adoption of the BRs where we finally agreed to a five year 
>>> timeline that would shift to 3 years in 2015.  I'm strongly opposed 
>>> to going back and re-evaluating these timelines now, especially 
>>> considering the amount of discussion, debate, and compromise went 
>>> into it last time.
>>>
>>> -----Original Message-----
>>> From: public-bounces at cabforum.org
>>> [mailto:public-bounces at cabforum.org]
>>> On Behalf Of Sigbjørn Vik
>>> Sent: Friday, August 02, 2013 4:56 AM
>>> To: public at cabforum.org
>>> Subject: Re: [cabfpub] Concerns regarding Mozilla Root 
>>> Program/Baseline Requirements
>>>
>>> I think the BRs are clear on what they require, and the EV 
>>> definitions clearly spell out that renewals and reissuances create 
>>> newly issued certificates. All new certificates require valid validity
periods.
>>> While I can understand that people may want to discuss if the rules 
>>> are optimal, I dislike CAs claiming ignorance of the rules, or 
>>> exception to the rules since they have done things differently in 
>>> the past. If someone wants to claim the rules are unclear, do so 
>>> before assuming a particular interpretation, not afterwards.
>>>
>>> Opera is happy to work with the group to define optimal rules, but 
>>> we dislike CAs engaging in creative reinterpretation of the rules. 
>>> If we can't trust that CAs follow the rules decided upon, then the 
>>> rules are close to worthless.
>>>
>>> In addition to reasons mentioned previously about why constrained 
>>> validity periods is a good idea, there is also the point about the 
>>> web moving forwards. For us to be able to upgrade requirements, 
>>> ditch insecure practices, and promise a secure experience in the 
>>> future, we need to limit the time limit of certificates today. This 
>>> is basic insurance for the web in the future. Any exceptions we 
>>> carve out today, take away from this insurance, and might end up 
>>> costing us a lot. This is not a zero-sum game, and decisions we make 
>>> today might have serious consequences some years down the road.
>>>
>>> This discussion has also highlighted some other issues:
>>> * It seems some CAs are happy to issue backdated certificates. We 
>>> should probably spell out that signing, issuance and first validity 
>>> dates should all be as close as possible, with a maximum discrepancy 
>>> allowance.
>>> * To be able to verify this, we should also require that any CT 
>>> registrations happen within that time limit. I will contact the CT 
>>> team and ask them to consider this.
>>> * It seems audits don't catch CAs issuing certificates contrary to 
>>> the BR.
>>> What can we do to ensure that audits catch such issues?
>>> * Does the CABForum need an explicit objective?
>>>
>>> On 01-Aug-13 15:08, Rich Smith wrote:
>>> > Taking Ryan's definition and subsequent logic would effectively 
>>> > nullify all pre-BR agreements made with thousands of CA customers 
>>> > who bought certificates for terms greater than the now allowed 60 
>>> > months max (and soon to be 39 month max).  My lay opinion is that 
>>> > such
>>> action
>>> > would open the CAs to lawsuits for breach of those agreements.
>>>
>>> For CAs not to follow the BRs would presumably open them to lawsuits 
>>> from those who rely on the CAs for trust. For browsers not to follow 
>>> due diligence and best practices would presumably open them to 
>>> lawsuits from their customers. So if you care about potential 
>>> lawsuits, the score is
>>> 2-1 for not issuing such certificates. However, I think potential 
>>> lawsuits are irrelevant. The goal of this group should be to secure 
>>> the web, not to ensure the short-term profits for individual members.
>>> (See the last point above, about an objective.)
>>>
>>> On 01-Aug-13 17:00, Steve Roylance wrote:
>>> > [...] in 2015 when the
>>> > industry flips to a 39 month max I believe from my engineering 
>>> > team
>>> we
>>> > would have had a similar issue to others, again due to the 
>>> > existing contract precedence understanding that others have mentioned.
>>>
>>> That would be strange. Since the BRs were adopted in November 2011, 
>>> all CAs have known (and been audited on) the fact that after 1 April
>>> 2015 the maximum validity period is 39 months. For CAs today to sell 
>>> certificates valid more than 39 months past that date, and promise 
>>> to reissue them at any point would be shooting themselves in the foot.
>>>
>>> --
>>> Sigbjørn Vik
>>> Opera Software
>>> _______________________________________________
>>> Public mailing list
>>> Public at cabforum.org
>>> https://cabforum.org/mailman/listinfo/public
>>>
>>> _______________________________________________
>>> Public mailing list
>>> Public at cabforum.org
>>> https://cabforum.org/mailman/listinfo/public
>>_______________________________________________
>>Public mailing list
>>Public at cabforum.org
>>https://cabforum.org/mailman/listinfo/public
>
>
> _______________________________________________
> Public mailing list
> Public at cabforum.org
> https://cabforum.org/mailman/listinfo/public
> <table class="TM_EMAIL_NOTICE"><tr><td><pre>
> TREND MICRO EMAIL NOTICE
> The information contained in this email and any attachments is 
> confidential and may be subject to copyright or other intellectual
property protection.
> If you are not the intended recipient, you are not authorized to use 
> or disclose this information, and we request that you notify us by 
> reply mail or telephone and delete the original message from your mail
system.
> </pre></td></tr></table>
>
> _______________________________________________
> Public mailing list
> Public at cabforum.org
> https://cabforum.org/mailman/listinfo/public




More information about the Public mailing list