Open Bug 2092247 Opened 4 days ago Updated 1 day ago

DigiCert: Certificate issued using a key previously demonstrated as compromised

Categories

(CA Program :: CA Certificate Compliance, task)

Tracking

(Not tracked)

ASSIGNED

People

(Reporter: dcbugzillaresponse, Assigned: dcbugzillaresponse)

Details

(Whiteboard: [ca-compliance] [__-misissuance])

Preliminary Incident Report

Summary

  • Incident description: DigiCert received a Certificate Problem Report from a community member that they had successfully requested a certificate whose private key corresponded to a certificate they had previously requested revoked for ‘keyCompromise’.

    DigiCert Support initially received a request by email from the reporter reporting a private key as compromised. The report demonstrated compromise by way of a message signed with the private key, rather than one of the demonstration formats described in our CP/CPS. It was not submitted through our Problem Report web tool as our CP/CPS directs.

    Nevertheless, DigiCert Support verified the signature, revoked the certificate with reason keyCompromise within 24 hours, and notified the subscriber. However, the public key was not added to our compromised key blocklist, because that step is only performed automatically for reports submitted through the Problem Report tool, and the reporter was not directed to that tool.

    The reporter subsequently requested and was issued a new certificate using the same key pair and filed the CPR. That certificate has now been revoked and the private key added to the blocklist.

  • Relevant policies: TLS Baseline Requirements v2.3.1:

    • Section 6.1.1.3, item 4 (Subscriber Key Pair Generation)
    • Section 4.9.1.1 (reason 3) (Reasons for Revoking a Subscriber Certificate)
    • Section 4.9.3 (Procedure for Revocation Request)
    • Section 4.9.12 (Special Requirements Related to Key Compromise)
    • DigiCert CPS v7.09, Section 1.5.2 (Contact Person) and Section 4.9.3 (Procedure for Revocation Request)
  • Source of incident disclosure: Third Party Reported

I hope the full report does not also sound like you're blaming the reporter for not manually adding it to the list for you.

Note: I'm not the reporter for this.

Also previous reports for reissuance of keyCompromise reported to DigiCert were dismissed, including mastercard.com which continues to occur. I hope the full incident report includes prior warnings, and covers all ongoing scenarios for DigiCert's divergence compared to other CAs on how to handle keyCompromise events.

I presume this is regarding datayard.us, however keep in mind: mastercard.com, ssltechs.net, automatetls.com, grapefruitdesk.com are also amongst this class of ongoing non-compliance.

As per your reply on October 8th:

We have investigated your two most recent reports relating to keyCompromise and disagree that there is an incident.

As you note, the TLS BR says in Section 6.1.1.3:

The CA SHALL reject a certificate request if one or more of the following conditions are met: …
(4) The CA has previously been notified that the Applicant’s Private Key has suffered a Key Compromise using the CA’s procedure for revocation request as described in Section 4.9.3 and Section 4.9.12....

Section 4.9.3 of the TLS BR in turn says:

The CA SHALL provide a process for Subscribers to request revocation of their own Certificates. The process MUST be described in the CA's Certificate Policy or Certification Practice Statement. The CA SHALL maintain a continuous 24x7 ability to accept and respond to revocation requests and Certificate Problem Reports.

The CA SHALL provide Subscribers, Relying Parties, Application Software Suppliers, and other third parties with clear instructions for reporting suspected Private Key Compromise, Certificate misuse, or other types of fraud, compromise, misuse, inappropriate conduct, or any other matter related to Certificates. The CA SHALL publicly disclose the instructions through a readily accessible online means and in Section 1.5.2 of their CPS.

Section 4.9.12 of the TLS BR merely points back to Section 4.9.3.

Section 4.9.12 of our CP CPS clearly states:

  • If the entity requesting revocation for keyCompromise can demonstrate possession of the certificate’s Private Key, then DigiCert will revoke all instances of that key across all subscribers.
  • If the entity requesting revocation cannot demonstrate possession of the certificate’s Private Key, then DigiCert may revoke all certificates associated with that subscriber that contain that Public Key.

Our approach was based on discussions within the CABF and in MDSP conversations that laid out this preferred approach. As discussed exhaustively in that conversation, there are important reasons why a CA should not be required to revoke all certificates where possession of the Private Key has not been demonstrated. But there are circumstances where revocation would be appropriate even without demonstrated possession. Following CABF recommendations, therefore, we provided that we “may” revoke associated certificates, even though that is not our standard practice.

We have investigated the logs for the certificates in your reports, and have confirmed that they were revoked by the Subscriber(s) via our Portal without demonstrating possession of the Private Key. In our Portal, our instructions note that we will only block onward use of the Private Key if possession is demonstrated using https://problemreport.digicert.com/.

To be clear, in all cases where we have demonstrated possession of the Private Key, we block future use. We confirm that this control is operational.

DigiCert has acted in compliance with our CP CPS, which is compliant with the relevant section of the TLS BRs. As such, based on the facts currently available, we appreciate your report, but do not plan to open an incident.

Then in your later reply:

As you point out, there does appear to be lingering inconsistency among CAs on keyCompromise revocations. The subject was discussed at length— and we understood was settled— on the MDSP list a few years ago. You can find that dialogue at https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/RVFnJJRuUag/m/JG2JLLbmBAAJ. As an outcome of those discussions, DigiCert and a number of other CAs adopted this approach.

We note that the CABF recently opened a discussion on simplifying TLS revocation codes; we will participate there and hope that lingering differences on this point can be settled.

Did you ever raise this concern to CABF and actually attempt to resolve this matter prior to this incident? I believe you were adequately forewarned. This also applies to Globalsign, who are the only other divergent CA on this matter but do not have active misissuances where DigiCert stands alone.

Assignee: nobody → dcbugzillaresponse
Status: UNCONFIRMED → ASSIGNED
Ever confirmed: true
Whiteboard: [ca-compliance] [__-misissuance]

In response to Comment 1:

DigiCert is not blaming the reporter. As described in the DigiCert CP/CPS and in our customer portal, we will automatically block onward use of the Private Key if possession is demonstrated using https://problemreport.digicert.com/. In this case, our support team manually confirmed the novel demonstration and did not take the additional manual step of adding the key to the blocklist. Our final incident report will describe the steps we are taking to prevent re-occurrence.

For other readers’ background, in October Wayne reported to DigiCert that he had identified certificates including keys that matched certificates that previously had been revoked for keyCompromise. His comment includes our response: the TLS BR purposefully do not require a CA to block keys when possession is not demonstrated during the revocation request. We stand by our policy, which was based on detailed discussions within the CABF and in MDSP that laid out this preferred approach. Additional significant CAs, beyond those mentioned by Wayne, follow this practice.

In this case, possession was demonstrated, albeit not by DigiCert’s documented channel, and the key should have been blocked.

The CABF discussion on revocation reason codes is proceeding.

You need to log in before you can comment on or make changes to this bug.