Skip to content
Blog

Red Hat · Quay · container registry

Red Hat Quay: Two Critical Flaws Expose Private Registry Images (CVE-2026-123710, CVE-2026-110697)

Red Hat Quay: an unauthenticated signing-key injection (9.8) and cross-tenant robot token theft (9.9). How they work, what's at stake, and what to do about it.

5 min readCVE-2026-123710CVE-2026-110697

On November 11, 2026, Red Hat published two critical vulnerabilities in Quay, its container image registry:

  • CVE-2026-123710 (9.8): an unauthenticated client can get the registry to accept its own signing key, then read and write any repository;
  • CVE-2026-110697 (9.9): a user in one tenant can retrieve robot accounts and their decrypted tokens from other tenants.

An image registry sits on the critical path of software delivery: whatever lands there ends up deployed in production. Being able to replace a tag in a private repository means getting your code executed by every cluster that pulls it.


The Two Vulnerabilities

FieldCVE-2026-123710CVE-2026-110697
CVSS 3.1 (Red Hat)9.8 (CRITICAL)9.9 (CRITICAL)
VectorAV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:HAV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H
TypeCWE-1287 (input type validation)SQL injection (CWE-89)
AuthenticationNoneAn account with a namespace
NVD published2026-11-112026-11-11
NVD statusAwaiting AnalysisAwaiting Analysis

The scores come from Red Hat, the CNA for these CVEs. NVD hasn't analyzed them yet and publishes no version configuration — no affected or fixed version in its data. To find out whether your version is affected, go to Red Hat's page for each CVE and the linked Bugzilla tickets (2568072 and 2552237).


CVE-2026-123710 — a signing key installed by a stranger

From NVD:

An unauthenticated client can obtain an anonymous registry token with an audience it controls, present that token to the service key rotation endpoint, and install its own approved signing key with no expiry.

Three steps:

  1. the client obtains an anonymous registry token, picking its audience itself;
  2. it presents that token to the service key rotation endpoint;
  3. Quay registers its signing key, approved and with no expiry.

From then on, any token signed with that key is accepted for any repository: pulling private images, pushing, replacing tags.

The design flaw is visible in the description itself. A token's audience field exists precisely to say which service it's meant for. If the client chooses the audience and the rotation endpoint doesn't check where the token came from, an anonymous token turns into an admin credential.

One point matters for remediation: the installed key never expires. Patching Quay may not be enough to close the door if a rogue key has already been registered — the approved service keys need reviewing too.


CVE-2026-110697 — one underscore to read your neighbours' robots

From NVD:

The robot account listing function constructs a SQL LIKE pattern from user-controlled namespace names without escaping wildcard metacharacters.

In a SQL LIKE pattern, _ means "any single character". Quay builds the pattern from the namespace name without escaping that wildcard. An attacker who creates a namespace containing an underscore gets a pattern that also matches other tenants' namespaces whose names differ by one character at that position.

The robot account list they get back then includes other people's, with their decrypted authentication tokens. Robot accounts are what CI/CD pipelines use to push and pull images, so their tokens carry the pipelines' rights.

The vector says it all: PR:L (you need an account that can create a namespace) and S:C (the impact escapes the tenant boundary). This one hurts most on shared instances, where different teams or customers use the same Quay.


Who Is Affected

NVD doesn't list affected versions. In practice:

  • Self-hosted Quay (Red Hat Quay, Quay deployments on OpenShift): check Red Hat's pages for both CVEs to see which versions are affected and which update to apply.
  • CVE-2026-123710 concerns any instance whose key rotation endpoint is reachable by unauthenticated clients.
  • CVE-2026-110697 is most critical where several organisations or teams that don't trust each other share the instance and can create namespaces.

Detection

  • Service keys: inventory approved keys, especially those without an expiry date or created recently without an administrator's action
  • Access logs: calls to the key rotation endpoint from anonymous clients
  • Modified tags: tags replaced in private repositories outside normal deployment windows, digests changing with no matching build
  • Namespaces containing _ recently created by little-known accounts
  • Robot accounts used from unusual addresses or pipelines

Mitigation

1. Update Quay

Apply the fixed version Red Hat lists for your branch.

2. Purge suspicious service keys

A key installed through CVE-2026-123710 stays valid after the update, since it's approved and never expires. Remove any key whose origin isn't documented.

3. Regenerate robot account tokens

On a shared instance, assume robot tokens may have leaked: regenerate them and update pipeline secrets.

4. Check image integrity

If you sign your images, verify signatures on deployed tags. If you don't, compare production digests against those your builds produced.

5. Until the fix is in

Reduce exposure: no anonymous access from the internet unless you need it, and restrict namespace creation to trusted users.


Why Continuous Monitoring Matters for Your Delivery Chain

An image registry rarely gets inventoried like a piece of equipment: it runs in a cluster, owned by the platform team, out of security's line of sight. Yet it's the one place where a single flaw lets someone change what runs everywhere else.

With cveo.tech, add your infrastructure components — registry included — to your inventory and get an alert as soon as a critical CVE affects them.

Know which CVEs hit the versions you run.

Add the product and its version: cveo.tech lists the CVEs that affect it right away, then alerts you by email when a new one affects it — including before NVD has finished its analysis. Free, an account is all it takes.

Every Monday

The week's critical CVEs, in your inbox

One email a week: the CVSS ≥ 9 vulnerabilities published in the last seven days, plus our latest analyses. Nothing else.

Double opt-in by email. Unsubscribe in one click, any time.