Thanks for this summary, Ben. WRT point 3, I think a good way to roll out randomly-sampled certificate revocation is as an amendment to the CPS of all CAs; CAs amend their CPSes all the time without major contractual upheaval, so it seems like a useful tool here. Coupled with an appropriate effective-after date, we could be sure that all issued certificates were governed by a CPS version that includes random revocation as audit mechanism.
WRT the random selection itself, I would like to see them chosen via an externally-transparent process rather than leaving it to CAs to be appropriately random. Whether by Mozilla or a CCADB process, something that samples from crt.sh in an appropriate way, with a public randomness seed, would give higher confidence in the process. CAs will otherwise have a strong incentive to “roll the dice again” if the random selection includes a major or stubborn customer, and those are the cases that most need to be tested IMO. I think for all of these proposals, it’s important to state what the consequences of violation will be. I don’t think that Mozilla should feel constrained to only have policies that exactly match the TLS BRs. They are *baseline* requirements, not exhaustive requirements, and each cert consumer has its own context in which to make tradeoffs between inclusiveness and rigour. More broadly, it would be good to state what the main purpose and values of the MRSP are: for example, is the goal to permit as many CAs as can meet the requirements to encourage broad competition, or to optimize for the security of Mozilla’s users (which might mean having fewer CAs for which oversight by Mozilla can be more thorough)? Mike On Wed, Dec 18, 2024 at 11:17 AM 'Ben Wilson' via [email protected] <[email protected]> wrote: > All, > > As part of the discussions on this proposal, namely that CAs “maintain and > test mass revocation plans annually, including the revocation of 30 > randomly chosen certificates within a 5-day period,” I’ve received a few > comments via private channels, and to ensure transparency and foster > discussion, I am sharing them here anonymously: > > 1. “Mozilla does not grant exceptions…” -- this is the most important > signal that Mozilla can provide. > > 2. If certificate consumers want to prohibit delayed revocation, then they > need to make it clear to CAs that they won't accept it and that they will > kick them out of the root stores if they still do it. Don't try to solve > this issue with indirect measures like random revocations. Just be straight > about it and make it clear that there will be consequences for the very > first delayed revocation and onward. > > 3. We will face big problems in revoking productive customer certificates > just to test our mass-revocation plan and procedures. Our current customer > contracts do not foresee this. While we can revoke at any time for security > or compliance reasons, this authorization should not be used just to test > mass-revocation. This will also require us to push out contract changes to > our complete TLS customer base, which will take a considerable amount of > time and effort. > > 4. This part of the proposal should occur within the CA/Browser Forum > through amendments to the TLS Baseline Requirements, and not via Mozilla > Root Store Policy. > > 5. Why was the number 30 chosen as a sample? Some CA operators issue very > few certificates, while some CAs issue millions of certificates. > > I welcome your feedback on these points, the random sampling proposal, and > any others. > > Thanks, > > Ben > On Monday, December 16, 2024 at 3:02:35 AM UTC-7 [email protected] wrote: > >> Hi Ben, >> About " annual plan testing by revoking 30 randomly chosen certificates >> within a 5-day timeframe; and"... >> We understand that this mean that a CA will need to randomly revoke 30 >> certificates that most likely don't have other reason for being revoked >> than being randomly chosen and customers will just need to "happily" accept >> the situation... Harsh, but doable... >> >> About "audit report submitted under section 3.1 SHALL include an >> attestation that the CA operator has met these mass revocation planning >> requirements", it must be considered that the attestation letters of >> Webtrust audit reports have a fixed format, so such addition would be added >> most likely as a "Other matters" section, that audits can take each >> differently. >> >> My question here would be if you think it's there any chance that these >> requirements become part of the TLS BRs instead of the Mozilla Policy, I >> see several benefits here: >> - Checking the mass-revocation plan would be integral part of the audit >> scope, so auditors don't need to figure out how to include it in the >> reports... it just needs to be added to the audit criteria. >> - The "inverse-lottery" thing could be added in the revocation timelines >> of the BR, so there's an entry in the 5-day deadline adding a new category >> "The certificate has been randomly chosen for revocation during an internal >> audit". This should facilitate the contractual language to add in the >> subscriber agreement. >> >> El domingo, 15 de diciembre de 2024 a las 21:51:43 UTC+1, Ben Wilson >> escribió: >> >>> All, >>> >>> The purpose of this email is to start discussion of Mozilla GitHub Issue >>> #276 <https://github.com/mozilla/pkipolicy/issues/276> ("Address >>> Delayed Revocation"). We would like to collect comments and feedback on a >>> proposal to address delayed certificate revocation from a Mozilla >>> perspective. It builds on prior discussions and feedback regarding delayed >>> revocation, and the proposal is meant to replace guidance currently >>> provided on the Mozilla CA wiki >>> <https://wiki.mozilla.org/CA/Responding_To_An_Incident#Revocation>. >>> >>> Here is the comparison link for a proposed new section 6.1.3 in the MRSP: >>> >>> >>> https://github.com/mozilla/pkipolicy/compare/51b2f702accd54cb70d52081a9e814298433495b%E2%80%A6efa8ac40ac341fb813620938ef72328a53858038 >>> >>> >>> *Summary* >>> >>> Here are the highlights of the proposal: >>> >>> >>> - Revocation must occur promptly in compliance with the timelines >>> set in section 4.9.1 of the TLS Baseline Requirements (TLS BRs). Mozilla >>> does not grant exceptions to these timelines. >>> - New CA Obligations: >>> - Educate subscribers on revocation timelines and discourage >>> reliance on certificates in systems that cannot tolerate timely >>> revocation. >>> - Include contractual language requiring subscriber cooperation >>> with revocation timelines. >>> - Maintain and test mass revocation plans annually, including the >>> revocation of 30 randomly chosen certificates within a 5-day period. >>> - Beginning April 15, 2026, CA audit reports must attest to >>> compliance with the mass revocation planning requirements. >>> - Delayed revocation incidents must be reported per version 2.1 of >>> the CCADB's Incident Reporting Guidelines (as currently proposed >>> <https://github.com/mozilla/www.ccadb.org/pull/187>) >>> - Repeated delayed revocation incidents will result in heightened >>> scrutiny or sanctions, which may include root removal. >>> >>> *Background* >>> >>> Earlier this year, on this list, I proposed an Interim Policy to >>> Address Delayed Revocation >>> <https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/hXr43W3c4Gs/m/J1OAktIaAwAJ>. >>> While the proposed interim policy provided clarity, it faced criticism >>> regarding implementation complexity, burden on subscribers and CAs, and the >>> feasibility of associated measures, such as transitioning delayed >>> revocation domains to 90-day certificates. Also, there were subsequent >>> proposals aimed at reducing certificate lifetimes and encouraging >>> automation. See e.g. https://github.com/cabforum/servercert/pull/553. >>> >>> This new proposal drops proposed measures such as domain-specific >>> tracking and subscriber attestations and instead focuses on subscriber >>> education, mass revocation preparedness, and robust incident reporting >>> as the primary mechanisms for improving agility and transparency regarding >>> delayed revocation. >>> >>> If adopted, the proposed MRSP § 6.1.3 would replace the current guidance >>> on delayed revocation in Mozilla’s wiki and ensure consistency with the >>> CCADB's Incident Reporting Guidelines. >>> >>> I welcome your feedback on this draft proposal. Please share your >>> thoughts, questions, or concerns to help us refine and improve it further. >>> >>> Thanks, >>> >>> Ben Wilson >>> >>> Mozilla Root Store >>> >> -- > You received this message because you are subscribed to the Google Groups " > [email protected]" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion visit > https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/71b07640-d425-4f2f-8da4-d97a9475b9f6n%40mozilla.org > <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/71b07640-d425-4f2f-8da4-d97a9475b9f6n%40mozilla.org?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "[email protected]" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/CADQzZqtJEvJ1DnqsdYxCFa%3D_urNJ1Zn_4dU5ot%3D%3D8AOBeJephg%40mail.gmail.com.
