Taiwan Information Security Association

Security alertsSat, 14 Nov 2026

Red Hat Quay vulnerabilities put registry signing keys and robot account tokens at risk

Two Red Hat Quay vulnerabilities can expose private images or allow unauthorized changes to images used by deployment pipelines. Restrict access to /keys, check the fix for your release, and review registry and CI/CD activity.

01

Article

Red Hat disclosed two critical Quay vulnerabilities on 11 November 2026. An attacker could gain unauthorized access to images or change what an application deployment pulls. Review shared enterprise registries and managed registries, including internal ones that untrusted users can reach.

  • CVE-2026-123710: Red Hat CNA CVSS v3.1 9.8. When an untrusted client can reach /keys, the service key rotation endpoint can accept a registry token as authorization to install an attacker-controlled signing key. With anonymous access enabled, this path requires no account. With anonymous access disabled, Red Hat says a valid account can still exploit it. The inserted key can remain trusted without an expiration date and can sign tokens for private repositories.
  • CVE-2026-110697: Red Hat CNA CVSS v3.1 9.9. A signed-in user who can create a namespace with an underscore that matches another namespace under SQL LIKE rules can retrieve that tenant’s robot account tokens. Access then depends on the token’s assigned rights. Red Hat also documents effects on a matched tenant’s robot accounts when the colliding namespace is deleted or, under separate privileges or configuration, renamed.

The two vulnerabilities affect different credentials. Replacing robot tokens or resetting passwords does not revoke an unauthorized service signing key. Affected robot tokens and trusted service keys need their own investigation and recovery steps.

Attack flowTwo routes from registry access to deployment risk
  1. 01

    Reach the relevant entry point

    One route requires access to /keys. The other requires a signed-in user who can create a matching namespace.

  2. 02

    Compromise registry trust

    Insert a trusted signing key or obtain another tenant’s robot token, depending on the vulnerability.

  3. 03

    Access or modify images

    Forge registry tokens or use stolen robot credentials to pull private images, push images or replace tags within the available access.

  4. 04

    Reach downstream deployments

    A changed image can affect builds or clusters that later pull it.

Affected products and Red Hat updates

Review standalone Quay, the Quay Operator on OpenShift, and the mirror registry for Red Hat OpenShift. Red Hat publishes branch-specific updates for Quay 3. Use the exact supported image and digest in the erratum for your deployment.

Product or serviceUpdate status
Red Hat Quay 3Red Hat associates the same eight branch errata with both CVEs: 3.9.28, 3.10.28, 3.12.24, 3.14.10, 3.15.9, 3.16.7, 3.17.7 and 3.18.3.
Red Hat Quay 5Red Hat lists quay/quay-rhel9 as affected by both CVEs. Its CVE product records did not list a Quay 5 fixed build as of 14 November. Check the latest entry for your product channel.
Mirror registry for Red Hat OpenShift 2Red Hat lists openshift/mirror-registry-rhel8 as affected by both CVEs. Check for a mirror-registry-specific update.
Quay.ioRed Hat reports that the hosted service has fixed CVE-2026-110697.
Response flowRecommended actions: from the perimeter to core systems
  1. 01

    Restrict the service key endpoint

    At the proxy or load balancer, allow only trusted internal components to reach /keys. Check direct routes to Quay and preserve the key rotation access legitimate components need.

    See what to doResources: 2

    TWISA member solutions

    Restrict the service key endpoint

    Review network access to /keys and validate that untrusted users cannot reach it through an alternate path.

  2. 02

    Inventory installations and downstream dependencies

    Find standalone installations, outsourced registries, OpenShift Operator deployments, mirror registries and dependent clusters. Record the actual branch, deployed image digest and owner, then match each installation to the Red Hat advisories.

    See what to doResources: 1

    TWISA member solutions

    Inventory installations and downstream dependencies

    Use asset and exposure management to match runtime registry images to the correct Red Hat release channel and track retesting.

  3. 03

    Deploy the fixed release

    Follow the matching Red Hat erratum. Roll out the fixed image and verify the image digest of every running registry instance.

    View guidanceResources: 2

    Response guidance

    Deploy the fixed release

    Follow the matching Red Hat erratum. Roll out the fixed image and verify the image digest of every running registry instance.

  4. 04

    Investigate signing keys and robot tokens separately

    Preserve service key and registry logs. For CVE-2026-123710, review service_key_rotate and service_key_delete events, including entries with a null expiration_date. Confirm the owner and purpose of approved keys with Red Hat’s procedure and remove unauthorized keys. For CVE-2026-110697, look for affected namespace collisions and unexpected robot account operations, then rotate relevant tokens and confirm legitimate jobs work.

    See what to doResources: 2

    TWISA member solutions

    Investigate signing keys and robot tokens separately

    Coordinate service key investigation separately from robot token rotation. Review what each credential could authorize and preserve evidence of unusual changes.

  5. 05

    Verify image integrity and recover from trusted artifacts

    Compare image digests and tag history with trusted build records. Trace any suspect image to the CI/CD jobs and clusters that used it. If a change was unauthorized, isolate affected artifacts, restore from verified builds, and retest registry access and deployments.

    See what to doResources: 2

    TWISA member solutions

    Verify image integrity and recover from trusted artifacts

    Trace images from build through deployment and identify where an unauthorized tag change or unexpected digest could have reached production.

An unpatched Quay instance can accept a new unauthorized key after you remove one. Keep access to /keys restricted during investigation and apply a fixed build. When a fix for CVE-2026-110697 is pending, restrict creation of organizations or namespaces containing underscores, rotate tokens in colliding namespaces and monitor unexpected pulls or pushes. These steps reduce exposure while you arrange the update.

Response flowAuthorized testing and retesting
  1. 01

    Discover assets and versions

    Find registry instances, management routes and image versions, and scan for affected deployments.

  2. 02

    Test authorization boundaries

    Use authorized penetration tests, API tests and attack simulations to check /keys exposure and tenant isolation in a controlled environment.

  3. 03

    Verify the request path

    If a WAF fronts the relevant HTTP route, confirm the request, alert, actual block and backend outcome individually.

  4. 04

    Retest after updates

    Verify tenant isolation, key access and normal deployments. Assess host privilege escalation and internal reachability separately.

Defense architectureProtect registry trust through to deployment

Map self-managed and outsourced registries, Operators, mirror registries, image digests and deployment pipelines.

Allow authorized traffic and apply controls at each layer.

  1. 01

    Network and identity boundary

    Proxy, load balancer and authorized services

    Restrict /keysAllow access from required trusted components.

  2. 02

    Quay and its credentials

    Registry, namespaces, signing keys and robot accounts

    Patch and investigateVerify the fixed build and review unauthorized keys and cross-tenant access.

  3. 03

    CI/CD and runtime

    Build pipelines, images and clusters

    Verify image integrityCompare digests and deployment history, and recover from trusted builds.

Feed logs and traffic data from each layer into monitoring and response.

  1. Correlate registry activity with deployment recordsCombine logs for key operations, robot accounts, registry pulls and pushes, and deployment changes.
    See what to doResources: 2

    TWISA member solutions

    Correlate registry activity with deployment records

    Compare image and service credential activity across the registry, CI/CD platform and runtime.

TWISA member solutions

Official sources

For further product and service details, please contact the relevant TWISA members.

Advisory updated: 2026/11/14. Check the vendor’s latest information for product support and fixes.