
DigiCert Closed the Hole on July 2. It Took Until October 12 to Check Whether Anyone Had Walked Through.
A February 2025 server migration quietly moved a customer-ID archive onto the open internet. DigiCert spotted the stray accounts in June, blocked them, and then didn't check what they'd actually touched for another three and a half months.
CVE Tools4 min read
On November 1, 2026, certificate authority DigiCert filed a full incident report on Mozilla's public CA-compliance Bugzilla tracker (bug 2084460). It describes a 16-month chain of events that started with a routine infrastructure migration and ended with a third party reading archived customer ID documents — passports, driver's licenses, utility bills — for five days in the summer of 2026.
What actually happened, in order
16 months, from migration to full report
- Internal archive system migrated off its VPN-restricted segmentA business system used to archive customer-submitted ID documents was moved to an internet-reachable availability zone. Per DigiCert's filing, the change was tracked as infrastructure work, not a security change — so it bypassed the review that would normally flag new external exposure.
- DigiCert spots unfamiliar accounts, calls them harmlessInternal staff noticed accounts that weren't assigned to any application and judged the anomaly low-risk. DigiCert deployed a WAF rule blocking new account creation the same day.
- DigiCert's own security research program raises the same flagA researcher inside DigiCert's security program flagged the possibility that the accounts had been created externally — three days after the first sighting.
- Unauthorized read access to the archive beginsDigiCert's investigation later determined the actual data-access window started here, using one of the unassigned accounts.
- The hole is patched — but not investigatedDigiCert deployed an authorization fix (unassigned accounts now get a 401) and the access window closed. At this point DigiCert still treated it as an account-creation event, not a data-access event.
- Policy fix and pen test, while the data question sits unaskedDigiCert adopted a mandatory security sign-off policy for future network changes and ran a third-party penetration test — process fixes, still with no active inquiry into what the June accounts had actually read.
- DigiCert first suspects the accounts read archived dataMore than three months after closing the vulnerability, DigiCert began investigating whether the blocked accounts had accessed customer records — not whether more could get in.
- Formal determination: this is a reportable eventDigiCert classified the incident under the disclosure rules that Mozilla, Google Chrome, Microsoft and Apple impose on certificate authorities.
- Full incident report filed on Mozilla's BugzillaBug 2084460 lays out the timeline, root cause, and a 13-item remediation checklist, two items still marked ongoing with a due date of November 6.
- Remediation checklist closed out; press coverage followsDigiCert told Recorded Future News it has no evidence the archived information has been misused and has referred the matter to law enforcement. The Register and others publish critical coverage of the 16-month timeline the same day.
What was actually exposed — and what wasn't
DigiCert has not published a count of affected customers or records; its filing and its customer notifications both describe the identity-document exposure as a "small proportion" of the archive. The split between the certificate authority's core trust function and the business system that leaked matters more than the headline:
The archive vs. the certificate authority
- Names, business email addresses, phone numbers, business addresses for a subset of customers
- For a smaller subset: identity documents — driver's licenses, passports, utility bills submitted for identity verification
- Read-only access only — the account could not write records
- CertCentral and DigiCert ONE platforms
- Certificate issuance, revocation, and validation workflows — zero certificates affected
- Customer login credentials to DigiCert's core services
The remediation checklist
| Action | Status | Date |
|---|---|---|
| WAF rule blocking the registration endpoint | Complete | 2026-06-18 |
| Authorization fix: 401 for unassigned accounts | Complete | 2026-07-02 |
| VPN-only restriction enforced at the WAF | Complete | 2026-07-09 |
| Mandatory security sign-off for network changes | Complete | 2026-08-28 |
| Third-party penetration test | Complete | 2026-08-31 |
| New file-access / request-volume alerting | Complete | 2026-10-15 |
| Password reset for all business-system users | Complete | 2026-10-17 |
| Self-registration feature fully deactivated | Complete | 2026-10-24 |
| Migration to private-IP load balancers | Complete | 2026-10-24 |
| Per-endpoint authorization / least-privilege review | Complete | 2026-10-30 |
| Audit: confirm no origin server is externally reachable | Marked ongoing in the filing | Due 2026-11-06 |
| Final review of account-creation alerting, with SOC escalation path | Marked ongoing in the filing | Due 2026-11-06 |
Thirteen items were originally listed; the table above collapses two early WAF/auth steps into one row each where DigiCert's filing describes them as a single change. DigiCert's November 6 statement to Recorded Future News said the checklist was complete, though the Bugzilla thread itself does not carry a dated comment closing out the last two items as cleanly as the press statement implies — worth checking the live ticket rather than any single news account, including this one.
FAQ
Does this affect my TLS certificates?
How many customers are affected?
Was this a sophisticated attack?
Has the exposed data been misused?
Why this keeps happening
The root cause DigiCert names in its own filing is mundane and common: a network change got classified as "infrastructure," not "security," and so it skipped the review gate that exists specifically to catch new external exposure. That classification gap is why a self-registration feature built for an unrelated use case ended up attached to an archive of ID documents with no MFA in front of it. It is the same failure mode behind most of the breach post-mortems we track — a control that existed, just not on the path the actual change took.
Sources: DigiCert incident report, filed on Mozilla's CA-compliance Bugzilla tracker (bug 2084460).