[Responding to the most recent message in the discussion, apologies if this causes unexpected threading.]
Hi all, Given the discussion thus far, I’d like to propose the following for the group’s consideration in an effort to help guide a second round of discussion (TBD, but expected to begin no earlier than September 30). *Objective 1: Enhance WHOIS/RDAP validation of gTLDs with comparable security properties to DNS-based validation and sunset WHOIS/RDAP for ccTLDs.* *Justification: * - A recent disclosure [1] demonstrated how threat actors could exploit deficiencies in the WHOIS protocol and WHOIS tools served via HTTPS websites to obtain fraudulent TLS certificates. - Discussions within the Mozilla Dev Security Policy (MDSP) community [2] further expressed corresponding risks related to WHOIS, while also noting that ccTLDs may not maintain accurate or up-to-date WHOIS server records. Several examples of inoperative WHOIS servers for ccTLDs were identified. - Discussion in [3] further called into question the reliability of ccTLD WHOIS servers given, per IANA, there is no global policy requirement for ccTLD managers to operate a WHOIS server, and if they do, what kind of information is provided. - Solutions to strengthen existing WHOIS lookup methods were proposed in [4] and are considered in this update. *Approach: * - Add the following requirements in Sections 3.2.2.4.2 (“Email, Fax, SMS, or Postal Mail to Domain Contact”), 3.2.2.4.12 (“Validating Applicant as a Domain Contact”), and 3.2.2.4.15 (“Phone Contact with Domain Contact”). “ *Effective January 15, 2025, when issuing Subscriber Certificates…* - *The CA MUST NOT rely on Domain Contact information obtained using an HTTPS website, regardless of whether previously obtained information is within the allowed reuse period.* - *The CA MUST NOT rely on Domain Contact information obtained using the WHOIS protocol (RFC 3912) or the Registry Data Access Protocol (RFC 7482) if the requested Domain Name contains a ccTLD, regardless of whether previously obtained information is within the allowed reuse period.* - *When obtaining Domain Contact information using the WHOIS protocol, the CA MUST query IANA's WHOIS server and follow referrals to the appropriate gTLD WHOIS server.* - *When obtaining Domain Contact information using the Registry Data Access Protocol, the CA MUST utilize IANA's bootstrap file to identify and query the correct RDAP server for the domain.* - *The CA SHOULD NOT rely on cached 1) WHOIS server information or 2) RDAP bootstrap data from IANA to ensure that it relies upon up-to-date and accurate information.*” *Objective 2: Sunset Methods 3.2.2.4.2 (“Email, Fax, SMS, or Postal Mail to Domain Contact”) and 3.2.2.4.15 (“Phone Contact with Domain Contact”).* *Justification:* - While solutions to strengthen WHOIS-relying DCV methods are considered in this update (above), there is limited public evidence of significant reliance on these methods, including in response to [2] and [5]. - Instead, discussion has identified at least one CA Owner has already sunset reliance on WHOIS [6], and another that has changed its approach [7] for relying on WHOIS since disclosure of [1]. - More modern and heavily relied-upon DCV methods offer advantages over the existing WHOIS-based methods, including greater opportunity for seamless certificate lifecycle management automation (e.g., [8] and [9]), while also benefiting from recently improved security practices [10]. These methods can also more effectively align subscriber capabilities with agility and resilience expectations necessary to respond to the revocation timelines described in the TLS BRs [11]. - Beyond the above, previous discussions within the CA/Browser Forum have raised concerns about the perceived value (e.g., [12]) and security (e.g., [13]) of the DCV methods relying on WHOIS, further supporting the rationale for their gradual sunset. *Approach:* - Add the following requirements to Sections 3.2.2.4.2 (“Email, Fax, SMS, or Postal Mail to Domain Contact”) and 3.2.2.4.15 (“Phone Contact with Domain Contact”). “*Effective September 15, 2025, the CA MUST NOT rely on this method.*” The effective dates considered in this update are intended to 1) address the immediate concerns identified by [1], 2) offer near-term and longer-term transition periods for subscribers and CA Owners relying on existing implementations of these methods, and 3) align with existing effective dates in the TLS BRs (e.g., [10]). The above proposed updates compared to the initial effort described in [14] are highlighted at [15]. A comparison of these proposed updates against the in-force BRs is available at [16] Comments are welcome either on-list or with suggested edits via GitHub (preferred) at [17]. Thanks, Ryan [1] https://labs.watchtowr.com/we-spent-20-to-achieve-rce-and-accidentally-became-the-admins-of-mobi/ [2] https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/FuOi_uhQB6U/m/hKJOz3XzAAAJ [3] https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/mAl9XjieSkA/m/oDNWxtPwAQAJ [4] https://archive.cabforum.org/pipermail/servercert-wg/2024-September/004839.html [5] https://archive.cabforum.org/pipermail/servercert-wg/2024-September/004844.html [6] https://aws.amazon.com/blogs/security/aws-certificate-manager-will-discontinue-whois-lookup-for-email-validated-certificates/ [7] https://bugzilla.mozilla.org/show_bug.cgi?id=1917896 [8] https://cabforum.org/working-groups/server/baseline-requirements/requirements/#32247-dns-change [9] https://cabforum.org/working-groups/server/baseline-requirements/requirements/#322419-agreed-upon-change-to-website---acme [10] https://cabforum.org/working-groups/server/baseline-requirements/requirements/#3229-multi-perspective-issuance-corroboration [11] https://cabforum.org/working-groups/server/baseline-requirements/requirements/#491-circumstances-for-revocation [12] https://archive.cabforum.org/pipermail/servercert-wg/2018-August/000113.html [13] https://lists.cabforum.org/pipermail/validation/2024-July/001995.html [14] https://archive.cabforum.org/pipermail/servercert-wg/2024-September/004825.html [15] https://github.com/ryancdickson/staging/compare/356799f0dcfe11deb0a375a11233403236ab72c9..7a2ea7b33611bebf006a99a9a82729f183143eac [16] https://github.com/ryancdickson/staging/compare/ba28d04894d69c8fac62850b9d0de5061658c7c5..7a2ea7b33611bebf006a99a9a82729f183143eac [17] https://github.com/ryancdickson/staging/pull/9 On Wed, Sep 18, 2024 at 3:11 PM Amir Omidi via Servercert-wg < [email protected]> wrote: > I do not know much about the state of subdomain auth deployment in the CA > ecosystem unfortunately. > > On Wed, Sep 18, 2024 at 2:09 PM Andrew Ayer <[email protected]> wrote: > >> Hi Amir, >> >> On Wed, 18 Sep 2024 15:48:38 +0000 >> Amir Omidi via Servercert-wg <[email protected]> wrote: >> >> > There are two CAs (Let's Encrypt and Google Trust Services) with >> > DNS-ACCOUNT-01 ( >> > https://datatracker.ietf.org/doc/draft-ietf-acme-scoped-dns-challenges/ >> ) >> > mostly ready to go. This draft is designed to solve the CNAME >> > delegation problem. >> >> It doesn't obviate the need to run an acme-dns server (or similar) but >> DNS-ACCOUNT-01 would indeed be a great help. Note that RFC9444 >> (subdomain auth) support is also needed as otherwise the subscriber >> has to add delegations for every hostname instead of just one per zone. >> Do you know what the state of CA adoption is there? >> >> In any case, I'll give this I-D a more thorough look and provide >> feedback in the ACME WG. >> >> Regards, >> Andrew >> > _______________________________________________ > Servercert-wg mailing list > [email protected] > https://lists.cabforum.org/mailman/listinfo/servercert-wg >
_______________________________________________ Servercert-wg mailing list [email protected] https://lists.cabforum.org/mailman/listinfo/servercert-wg
