[Servercert-wg] [EXTERNAL] Re: SC-59 Weak Key Guidance

Aaron Gable aaron at letsencrypt.org
Thu Jul 6 20:54:58 UTC 2023


We agree that, if we are going to have requirements for weak key checking,
they should be as clear as possible. However, we believe that overall we
should be moving away from weak-key requirements. Weak key checks do little
to protect websites: if their software is generating weak keys, it is going
to continue to generate weak keys, and end up with no certificate at all.
And CAs are already required to revoke and subsequently refuse to issue for
known-compromised keys.

Adding more clarity now would require CAs to engage in significant amounts
of due diligence (for example, to prove beyond a doubt that
https://github.com/titanous/rocacheck behaves identically to the
soon-to-be-blessed https://github.com/crocs-muni/roca). We would prefer not
to engage in that when weak key requirements may be fully removed shortly
thereafter.

On Thu, Jun 15, 2023 at 10:03 AM Clint Wilson <clintw at apple.com> wrote:

> I broadly agree that Debian weak key checking has had diminishing returns
> in ecosystem value over time, and think dropping those checks specifically
> to be a reasonable proposal, however I’m not supportive of dropping all
> weak key checks (especially as new vulnerabilities arise and are
> identified).
>
> Approach-wise, I would prefer moving forward with this ballot as-is, with
> a follow-on ballot to propose pulling out Debian weak key checks. Just my 2
> cents.
>
> Cheers,
> -Clint
>
> On Jun 8, 2023, at 7:14 AM, Tim Hollebeek via Servercert-wg <
> servercert-wg at cabforum.org> wrote:
>
> Yes, if the consensus is to drop them as a requirement, we aren’t
> necessarily opposed, as that also improves the clarity.  We would probably
> continue to do most of the checks, as they do provide value to our
> customers, but other CAs need not be required to do absolutely everything
> we do.
>
> -Tim
>
> *From:* Servercert-wg <servercert-wg-bounces at cabforum.org> *On Behalf Of *Wayne
> Thayer via Servercert-wg
> *Sent:* Tuesday, June 6, 2023 3:44 PM
> *To:* Aaron Gable <aaron at letsencrypt.org>; CA/B Forum Server Certificate
> WG Public Discussion List <servercert-wg at cabforum.org>
> *Subject:* Re: [Servercert-wg] [EXTERNAL] Re: SC-59 Weak Key Guidance
>
> I agree with Tim's comments about the need to clarify these requirements.
> I'm not opposed to dropping them, but if we do that, we need to remove BR
> 4.9.1.1 (4). If CAs are not required to check for weak keys, they should
> also not be required to revoke certs containing weak keys.
>
> Thanks,
>
> Wayne
>
> On Tue, Jun 6, 2023 at 11:42 AM Aaron Gable via Servercert-wg <
> servercert-wg at cabforum.org> wrote:
>
> I concur with the above: the effort required for weak key checking
> (especially for Debian weak keys, which requires generating and storing
> huge collections of keys, and doing so again every time the CA's potential
> key-space expands) outweighs the benefits. All of the weak-key incidents
> that I personally recall have only found keys purposefully generated by
> security researchers.
>
> Additionally, under the ACME protocol, anyone who can derive a private key
> from a public key can cause any certs using that public key to be revoked.
> So security researchers who do decide to continue scanning for various
> forms of weak keys can simply revoke the affected certs immediately.
>
> Aaron
>
> On Fri, Jun 2, 2023 at 11:22 AM David Kluge via Servercert-wg <
> servercert-wg at cabforum.org> wrote:
>
> GTS complies with the ballot as it is written, but we believe that we
> should carefully assess whether each required weak key check is a valuable
> use of resources. In particular, the requirement to check for Debian Weak
> keys appears to have little value relative to the cost to us. The Debian
> Weak key vulnerability was fixed in 2008, over 15 years ago. Any client
> that is submitting a vulnerable key is running unpatched code from 2008 or
> earlier and they are going to be vulnerable to a significant number of
> other critical CVEs merely from running unpatched code.
>
> The cost of maintaining this check is not free. The recent publication of
> ECDSA keys and atypical architectures meant that many CAs likely had to
> re-architect their Debian Weak key checks to accommodate the new
> expectations [1]. There is of course always maintenance of code even if the
> architecture is constant.
>
> As Ryan mentioned, it may be more prudent for CAs to spend their time on
> other areas. Ryan mentioned linting, but there are of course many other
> areas the ecosystem can improve upon. Strengthening domain validation,
> automation technologies for renewal, increasing capacity/availability to
> enable shorter-lived certificates, preparing for mass revocation scenarios,
> etc.
>
> [1] https://bugzilla.mozilla.org/show_bug.cgi?id=1789521
> <https://url.avanan.click/v2/___https:/bugzilla.mozilla.org/show_bug.cgi?id=1789521___.YXAzOmRpZ2ljZXJ0OmE6bzozMzA2NTI2NDU3ZDMyZWJhMjM0NzE0MTJkMGE0MzQyYjo2OjZhZjc6NDkyOTQzMTdjYjdmZGJlN2ExZTNiYzUyZGNkMTVhMGRhN2EzZjk4MTQwZWZmNWEzNjJhNjEzMzI1OTdmYWM0YjpoOkY>
>
>
> On Thu, Jun 1, 2023 at 7:58 PM Tim Hollebeek via Servercert-wg <
> servercert-wg at cabforum.org> wrote:
>
> Having these requirements does make some sense, as customers do submit
> these weak keys from time to time (though rather rarely …).  Where simple
> checks are able to detect them, it makes sense to do so.
>
> For us, there are basically two things that take up the vast majority of
> time related to weak keys:
>
>
>    1. The lack of clarity in the current requirements wastes a lot of
>    people’s time figuring out exactly what is required and what isn’t.  The
>    requirements as they exist today, with extremely vague and open-ended
>    guidance, are a compliance nightmare, taking far more time and effort than
>    the benefit they provide.
>    2. This ballot, and analyzing its many iterations over the years.  It
>    does provide additional clarity, and can and should fix issue #1.  IMO
>    there were, at times, too many efforts to make it perfect instead of good.
>    We should get it to good, pass it, and move on.
>
>
> I’m HIGHLY sympathetic to the suggestion that we’ve spent far more time on
> this than it was worth, but having spent that time, the additional clarity
> is actually useful.  It does occasionally prevent customers from getting a
> certificate with a key that is trivially weak, possibly with disastrous
> consequences.  And it’s pretty easy to comply with an explicit list of
> concrete technical checks that should be performed on a key before
> accepting it.
>
> -Tim
>
> *From:* Servercert-wg <servercert-wg-bounces at cabforum.org> *On Behalf Of *Ryan
> Dickson via Servercert-wg
> *Sent:* Thursday, June 1, 2023 10:32 AM
> *To:* Bruce Morton <bruce.morton at entrust.com>
> *Cc:* CA/B Forum Server Certificate WG Public Discussion List <
> servercert-wg at cabforum.org>
> *Subject:* Re: [Servercert-wg] [EXTERNAL] Re: SC-59 Weak Key Guidance
>
> Hi Bruce,
>
> Sorry for not being clear.
>
> I was using linting as an example of an automated check that would occur
> at roughly the same frequency as we observe necessary for weak-key checks
> (i.e., once for every certificate). While I can't easily quantify the
> difference in effort comparing linting versus weak-key checking, based on
> incident disclosures to Bugzilla - we often see issues that could have been
> prevented with linting, but rarely see incidents related to weak-keys. It's
> also not clear, on average, what % of certificate requests are rejected due
> to a violation of 6.1.1.3 (i.e., how prevalent is the "weak-keys problem"?)
>
> I didn't intend for us to shift the scope of this ballot to focus on
> linting, but instead, to understand whether CAs thought continued weak-key
> checking was considered valuable.
>
> Thanks,
> Ryan
>
> On Thu, Jun 1, 2023 at 10:09 AM Bruce Morton <Bruce.Morton at entrust.com>
> wrote:
>
> Hi Ryan,
>
> I like your direction, but I need some help understanding how “requiring
> pre-/post-issuance linting instead of weak-key checks” would reduce the
> effort by the CA? I’m assuming a CA can meet the proposed ballot by doing
> pre-issuance linting of the CSR?
>
>
> Thanks, Bruce.
>
> *From:* Servercert-wg <servercert-wg-bounces at cabforum.org> *On Behalf Of *Ryan
> Dickson via Servercert-wg
> *Sent:* Thursday, June 1, 2023 9:44 AM
> *To:* Dimitris Zacharopoulos (HARICA) <dzacharo at harica.gr>; CA/B Forum
> Server Certificate WG Public Discussion List <servercert-wg at cabforum.org>
> *Subject:* [EXTERNAL] Re: [Servercert-wg] SC-59 Weak Key Guidance
>
> WARNING: This email originated outside of Entrust.
> DO NOT CLICK links or attachments unless you trust the sender and know the
> content is safe.
> ------------------------------
> [back to discussing the ballot]
>
> Hi all,
>
> I raised the following question during the January 19th
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__https:/cabforum.org/2023/01/19/2023-01-19-minutes-of-the-server-certificate-working-group/__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIEn9o8sx8A$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OjU1NzM6ODVjN2M4MTM0NTdjZWY4MDY1ZDVmNTA3YzEwMDhmMTk0ZmYyOGZhZDljZTZkNDRlYzkyYzM3ZDg5NmExMjYzMTpoOkY>
>  SCWG meeting, but I recognize only some group members can participate in
> our regularly scheduled meetings.
>
> Do participants in this Forum feel that weak-key checks should be removed
> from the scope of a CA’s set of mandatory responsibilities?
>
> While I appreciate the work carried out by ecosystem members to produce
> this ballot, primarily led by the SSL.com
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__http:/ssl.com__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIElUnEh06Q$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OjQ3ZTU6ZDZjZjFlNmE1M2QwYzJiY2Y5NGNkNjU0MjdlYzkwMjY2YTg4OTUyYWRjZDBjNmQwN2UyNzBlMTY1ZTc3Nzc3MDpoOkY>
>  team, I struggle with the demonstrated security value of these checks
> compared to the overall effort they represent.
>
> In a recent Validation Subcommittee meeting where we focused on delegating
> parts of the domain validation process, we discussed that subscribers often
> make security decisions that can have considerable consequences but are
> ultimately beyond the CA’s scope of responsibility (for example, delegating
> domain validation to an insecure third-party service). Wouldn’t we consider
> using an outdated software application/library to generate key-pairs along
> the same lines?
>
> Beyond perceived security value, I also struggle with the opportunity cost
> of time spent evaluating weak keys and responding to weak key incidents. It
> seems to me that something like requiring pre-/post-issuance linting
> instead of weak-key checks is a better tradeoff and would be more valuable
> for the ecosystem (e.g., reducing the likelihood of unexpected customer
> impact due to prescribed revocations timelines in the BRs related to
> mis-issuance).
>
> As this is now in discussion, I wanted to again offer the perspective that
> maybe weak-key checks should not be in scope of a CA’s responsibilities in
> case others share the same opinion.
>
> - Ryan
>
>
> On Mon, May 29, 2023 at 1:18 AM Dimitris Zacharopoulos (HARICA) via
> Servercert-wg <servercert-wg at cabforum.org> wrote:
>
> Hi Clint,
> On 26/5/2023 6:45 μ.μ., Clint Wilson wrote:
>
> Hi Tom, Dimitris,
>
> I continue to be opposed to the SCWG trying to limit effective dates to 2
> per year. I think it’s entirely reasonable to align on a day of the month
> (I think the 15th has broadly been the only one I’ve heard proposed). I
> think it’s reasonable to try to avoid January and December. I also think
> there may be value in trying to reduce the overall number of effective
> dates somewhat. The dates I’m personally in favor of aligning on are
> February, April, June, August, and October 15th.
>
> If there’s a particular penchant towards March and September, however,
> then I’d be unopposed to March, May, July, September, and November 15th.
>
> For this ballot in particular, I think October 15 or November 15 2023 are
> feasible targets for implementing these changes and would greatly prefer
> closing this issue (open now for *more than 3 years*) sooner than later,
> especially given the number of incidents we’ve seen in the last years
> related to weak key vulnerabilities and CAs issuing certificates with weak
> keys.
>
>
> It's fine for me also to close this issue sooner than later which is why I
> recommended even the September 15, 2023 effective date.
>
> On the 2 document releases per year issue, this is a preliminary result
> after having long discussions. I was not aware of any opposition until now,
> but perhaps your opposition didn't consider the emergency options of the
> proposal? The "standardized release cycle for Guidelines" proposal
> addresses a series of concerns about the frequency and number of document
> updates, as highlighted in the presentation shared in my previous reply. If
> you recall, the proposal still allows the release of "Emergency Guidelines"
> that bypasses the 6-month regular release cycle. We still need to work on
> the details which I hope to make progress on after passing the first Bylaws
> updates that are already prepared, but I'm confident that all concerns will
> be addressed.
>
> If we use this ballot as an example for applying the "standardized release
> cycle for Guidelines", Apple would propose that this is an Emergency
> Guideline and specify an effective date that would not be one of March 15
> or September 15. If there was no opposition, we would proceed with a ballot
> that would result in an emergency guideline release and the proposed
> effective date exactly as we normally do today.
>
> I plan to start a separate thread to continue this discussion at the Forum
> level after we make some progress with the recently proposed Bylaws changes.
>
>
> Thanks,
> Dimitris.
>
>
> Thanks,
> -Clint
>
>
>
> On May 26, 2023, at 7:37 AM, Tom Zermeno via Servercert-wg
> <servercert-wg at cabforum.org> <servercert-wg at cabforum.org> wrote:
>
> Hello Dimitris,
>
> Thank you for the input.  We feel that September 15th does not provide
> enough time for CAs to implement these changes, but we are not against the
> March 15,  2024 effective date, if there is consensus from the Community.
>
> Thank you,
>
> Tom
> SSL.com
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__http:/ssl.com/__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIEmQZfKH8A$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OmE5ODI6Y2RjOGZmZGQyMWQwZDdlNDQ4ZDhlOTc5YWQ4MzMzMjhhNDJiODYyZTgwNjE2ZmYyZGUxMTAwZTQ2M2Q1YjhiZTpoOkY>
>
> *From:* Servercert-wg <servercert-wg-bounces at cabforum.org> *On Behalf Of *Dimitris
> Zacharopoulos (HARICA) via Servercert-wg
> *Sent:* Friday, May 26, 2023 1:54 AM
> *To:* servercert-wg at cabforum.org
> *Subject:* Re: [Servercert-wg] SC-59 Weak Key Guidance
>
>
>
> Hi Tom,
>
> Historically, the SCWG has been trying to avoid effective dates during
> January or December. I recommend using September 15, 2023 or March 15, 2024
> as possible effective dates. These two dates seem to be more favorable
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__https:/docs.google.com/presentation/d/1oTGVYqggQpQMR4Lktbu_L6DhuBVJzeuiFGd9EAU1zsE__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIEmlvDMTQg$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OmIwNmU6MjY1N2MzM2E2YTgwNGI3N2M3OWI1NzAzNDFlOGZhODYwZjFlY2Y0YjI2NGVjNDY3YWQzYmUzYWEwYjdlMDEzODpoOkY> than
> others.
>
>
> Thanks,
> Dimitris.
> On 25/5/2023 10:51 μ.μ., Tom Zermeno via Servercert-wg wrote:
>
> Purpose of Ballot SC-059 V3
>
> Several events within the community have led to concerns that the Baseline
> Requirements for the Issuance and Management of Publicly-Trusted
> Certificates (BRs) lacked a specificity required to properly guide CAs on
> matters dealing with the identification and processing of digital
> certificates based on private keys considered weak, or easy to ascertain.
> In the hopes that elaboration and clarity on the subject would be
> beneficial to the community, we are presenting updates to §4.9.1.1(“Reasons
> for Revoking a Subscriber Certificate) and §6.1.1.3 (Subscriber Key Pair
> Generation) of the BRs.
>
> The first update is to §4.9.1.1 and is made to expand the scope of easily
> computable Private Keys from “Debian weak keys” to “those listed in section
> 6.1.1.3(5)”.  While the initial language in the BRs did not exclude other
> concerns, the use of a single example could be interpreted to mean that
> other easily computable Private Keys are few and far between.  The next
> update was to §6.1.1.3(5), wherein we added specific actions to be taken
> for ROCA vulnerability, Debian weak keys - both RSA and ECDSA – and Close
> Primes vulnerability.  We also added a link to suggested tools to be used
> for checking weak keys. Finally, an implementation date of December 1, 2023
> was added to allow CAs time to update processes to meet the requirements.
>
> The following motion has been proposed by Thomas Zermeno of SSL.com
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__http:/ssl.com/__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIEmQZfKH8A$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OjQ1YmM6ZTY5NzU0YjkzNjYwNmY0ODY4Y2U2YTE5NDZkNzMxZGNmZmUxOTI1NmI4MjEwZmNlMWQ1OWI3ODA2YTY1MTY2MzpoOkY> and
> endorsed by Ben Wilson of Mozilla and Martijn Katerbarg of Sectigo.
>
> --Motion Begins—
>
> This ballot is intended to clarify CA responsibilities regarding weak key
> vulnerabilities, including specific guidance for Debian weak key, ROCA and
> Close Primes attack vulnerabilities, and modifies the “Baseline
> Requirements for the Issuance and Management of Publicly-Trusted
> Certificates” as follows, based on Version 2.0.0.
>
> Notes: Upon beginning discussion for SC-59, the then-current version of
> the BRs was 1.8.4; since that time several ballots have been approved,
> leading to the increment of the version to 1.8.7 and eventually 2.0.0,
> which is the latest approved version of the BRs.  The changes introduced in
> SC-59 do not conflict with any of the recent ballots. As observed with
> other ballots in the past, minor administrative updates must be made to the
> proposed ballot text before publication such that the appropriate Version #
> and Change History are accurately represented (e.g., to indicate these
> changes will be represented in Version 2.0.1).
>
>
>
> MODIFY the Baseline Requirements as specified in the following Redline:
> https://github.com/cabforum/servercert/compare/a0360b61e73476959220dc328e3b68d0224fa0b3...SSLcom:servercert:3b0c6de32595d02fbd96762cda98cdc88addef00
> <https://url.avanan.click/v2/___https:/urldefense.com/v3/__https:/github.com/cabforum/servercert/compare/a0360b61e73476959220dc328e3b68d0224fa0b3...SSLcom:servercert:3b0c6de32595d02fbd96762cda98cdc88addef00__;!!FJ-Y8qCqXTj2!ZiS9e88ZglROzMqbZ57HX5kGTEoEo89sE6TPRs6_RvRCnQOeD9zdbxklqxVQD2dlDqEbV24CVdHmWw9BkoRWIEkXOeodFw$___.YXAzOmRpZ2ljZXJ0OmE6bzo0YjI4MjA3MmQyM2YzYTA3YjI4ZjdjNDM2MGNmNjMzMzo2OjMzMDU6ODViNTk3NjgzZmUyNTQ2ZWE1YzMzZjE3ODk2MWE3NWZmYTdjZWRhMWUzZmM4YTE1YzM3YjE2YzZmYjc4ZDczYzpoOkY>
>
>
>
>
> --Motion Ends—
>
> This ballot proposes a Final Maintenance Guideline. The procedure for
> approval of this ballot is as follows:
>
> Discussion (11+ days) • Start time: 2023-05-25 19:00:00 UTC • End time:
> 2023-06-08 18:59:00 UTC
> Vote for approval (7 days) • Start time: TBD • End time: TBD
>
>
>
>
>
>
> _______________________________________________
> Servercert-wg mailing list
> Servercert-wg at cabforum.org
> https://lists.cabforum.org/mailman/listinfo/servercert-wg
> <https://url.avanan.click/v2/___https:/lists.cabforum.org/mailman/listinfo/servercert-wg___.YXAzOmRpZ2ljZXJ0OmE6bzozMzA2NTI2NDU3ZDMyZWJhMjM0NzE0MTJkMGE0MzQyYjo2OjI1MjI6OGRjZWVkNDU1MjJiOWY1NDY5MDNhYmMyMzI2ZGEyOWM0YTg2NDM2MTU0OTJiNzA4NTcyOWI0YjU2ZTliMDU0MzpoOkY>
>
> _______________________________________________
> Servercert-wg mailing list
> Servercert-wg at cabforum.org
> https://lists.cabforum.org/mailman/listinfo/servercert-wg
> <https://url.avanan.click/v2/___https:/lists.cabforum.org/mailman/listinfo/servercert-wg___.YXAzOmRpZ2ljZXJ0OmE6bzozMzA2NTI2NDU3ZDMyZWJhMjM0NzE0MTJkMGE0MzQyYjo2OjdmM2Y6MTc2OWM0YTdiMzRhZmFjNGZhZTRhN2Q3M2Q2NDZlYzYyYmMzOWM1MWJlNzhlNGJlNGMxOTVlNjM2OTZkMTFiYjpoOkY>
>
> _______________________________________________
> Servercert-wg mailing list
> Servercert-wg at cabforum.org
> https://lists.cabforum.org/mailman/listinfo/servercert-wg
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cabforum.org/pipermail/servercert-wg/attachments/20230706/6beebe6f/attachment-0001.html>


More information about the Servercert-wg mailing list