The short version
- The flaw. CVE-2026-123710, rated 9.8, is in the service that manages Red Hat Quay's token-signing keys. An attacker can add a signing key of their own, with no expiry date, and Quay then accepts tokens signed with it for any repository.
- Exposure. Red Hat lists every Quay 3 and Quay 5 version as affected, including standalone installs, Quay deployed by the Quay Operator on OpenShift and the mirror registry for OpenShift. Any instance untrusted clients can reach is exposed, internal registries included.
- Second flaw. CVE-2026-110697, rated 9.9, was disclosed at the same time and involves the robot-account listing code. A user who can create a namespace containing an underscore can match another tenant's namespace and obtain its robot tokens.
- Fix status. The Hacker News reports fixed builds for eight Quay 3 branches, while SecurityWeek says neither the advisory nor the bug report names a fixed version. On 12 November the CVE entry for CVE-2026-123710 still listed every Quay package as affected.
- First moves. Red Hat advises blocking the /keys endpoint at the proxy or load balancer, searching audit logs for unexplained key rotation events and installing a fixed release, because a deleted attacker key can simply be re-added.
Timeline
| Date | Event |
|---|---|
| 11 Nov | Fixes for CVE-2026-123710 and CVE-2026-110697 are merged into the Quay GitHub repository. |
| 11 Nov | Red Hat says anonymous access being switched off does not close the hole, because any valid Quay account can still exploit it. |
| 12 Nov | The flaws are reported publicly; checks that day show the CVE entry for CVE-2026-123710 listing every Quay package as affected with no fix. |
How the key-signing flaw works
Quay signs the tokens that permit image pushes and pulls with keys it trusts. The flaw lets an attacker add a key of their own to that trusted set with no expiry date, after which Quay accepts tokens signed with it for any repository. The attacker can retrieve private images and any secrets or credentials stored inside them, and can push or overwrite images that other systems then pull.
The key server verifies that a key-replacement request carries a token signed by a trusted key with the expected audience and issuer, but does not check that the token was issued for key rotation. An anonymous token from Quay's /v2/auth endpoint passes, because Quay signs it and the caller picks the audience. A planted key survives restarts and can also be used to delete the keys running instances rely on, so normal pushes and pulls start failing.
Red Hat's Dave O'Connor found both flaws. According to the bug report, the logic behind CVE-2026-123710 has been in Quay since its first public commit. Red Hat says deleting a planted key buys little time, since an attacker can add another until a fixed release is running.
The second flaw and who is affected
CVE-2026-110697 (9.9) sits in the code that lists robot accounts. It builds an SQL LIKE pattern from the namespace name without escaping wildcard characters, and in SQL LIKE an underscore matches any single character. A namespace with an underscore in the right place can therefore return another tenant's robot accounts with their decrypted tokens. No elevated privileges are needed and the victim does nothing. Where Quay is the integrated registry for OpenShift, stolen robot credentials could be used to inject malicious images into CI/CD pipelines.
Red Hat lists every version of Quay 3 and Quay 5 as affected by both flaws. For CVE-2026-123710 its guidance names standalone installs, Quay deployed by the Quay Operator on OpenShift, and the mirror registry for Red Hat OpenShift. Any Quay instance that untrusted clients can reach is exposed, internal registries included. Quay enables anonymous access by default, and Red Hat's guidance is blunt: "Turning off anonymous access doesn't protect you."
Fix status and the paper trail
The sources differ on fixes. SecurityWeek reports that neither the advisory nor the bug report names a fixed version, though fixes for both flaws were merged into the project's GitHub repository on 11 November. The Hacker News reports that Red Hat released fixed builds for eight Quay 3 branches, and that the company's CVE page names the same eight releases as fixed for CVE-2026-110697 while none of the eight advisories list that CVE.
When The Hacker News checked on 12 November, the CVE entry for CVE-2026-123710 still showed every Quay package as affected with no fix, even though each advisory lists the flaw. For CVE-2026-110697, Red Hat says Quay.io is already fixed. The upstream change adds the missing rotation check, requires keys added through the key server to carry an expiry date, stops callers choosing a token's audience, and fixes the robot-account search.
What is not known yet
- Whether either flaw has been exploited against a real registry.
- How many Quay instances are reachable by untrusted clients and therefore exposed.
- Whether private images, embedded credentials or CI/CD pipelines have actually been read or tampered with.
- Whether fixed Quay 5 builds are available, since Red Hat lists Quay 5 as affected.
- Why the eight advisories do not list CVE-2026-110697 even though the company's CVE page treats it as fixed.
What to do
If you run Red Hat Quay
- Block the /keys endpoint at your proxy or load balancer for everything except trusted internal components.
- Search audit logs for service_key_rotate and service_key_delete events where the expiration date is null.
- Remove any approved service keys you do not recognise.
- Install a fixed release; deleting a planted key on its own does not last, because it can be re-added.
If you create namespaces or use robot accounts
- Limit who can create organisations or namespaces whose names contain an underscore.
- Rotate robot tokens for any namespace that matches an underscore namespace.
- Watch for pulls and pushes that do not fit normal robot account use.
For everyone else
- Do not treat disabling anonymous access as a fix: any valid account can still exploit the key-signing flaw.
- Check which key-management and administrative endpoints your registries expose to untrusted networks.
Update log
- First published.
Sources
New incident reports by email
One email a week with the reports published since the last one.