On this page (6 sections)

A critical flaw tracked as CVE-2026-110697 affects Quay. According to the NVD entry, the function that lists robot accounts builds a SQL LIKE pattern from namespace names without escaping wildcard characters. A user who creates a namespace containing an underscore can match robot accounts in other tenants' namespaces and retrieve their decrypted authentication tokens.

If you run a shared or multi-tenant Quay registry, or use Quay robot accounts in CI/CD pipelines, this deserves prompt attention. The public record is thin at the moment, so this article separates what is confirmed from what you will need to verify with your vendor.

What happened

The NVD description says the robot account listing code uses namespace names supplied by users when it builds a database search pattern. Wildcard characters in that pattern are not escaped. The result, per the description, is that an attacker can match and retrieve robot accounts, and their decrypted authentication tokens, from other tenants' namespaces.

The record points to Red Hat's CVE page and a Bugzilla entry as references. Only one independent publisher covers the story so far, and the sources available to us do not list affected versions, fixed versions or a patch. Treat any version claim you see elsewhere as unverified until you confirm it with the vendor advisory.

How the weakness works

The following is general background on SQL, not a detail taken from the sources. In a SQL LIKE comparison, some characters act as wildcards: an underscore matches any single character, and a percent sign matches any run of characters. When an application puts user input into a LIKE pattern without escaping those characters, the input stops being a literal name and becomes a loose search rule.

Here, the NVD description says a namespace name containing an underscore can be used to match robot accounts in other tenants' namespaces. This is not the classic case of injecting arbitrary SQL commands, but it is filed under CWE-89 because user-controlled input shapes a query without proper neutralization.

The impact is serious because robot account tokens are authentication credentials. The NVD description names cross-tenant credential theft and unauthorized image operations as the outcome.

Who is realistically exposed

  • Shared Quay registries where users can create their own namespaces. The description says the attacker creates a namespace with an underscore in its name, so this is the scenario it depends on.
  • Organizations whose CI/CD pipelines, deployment tools or clusters use Quay robot accounts, since those are the credentials at risk.
  • Teams that give robot accounts write access to production repositories. As general reasoning, not a statement from the sources, a stolen token with write access could be used to alter images, which matches the 'unauthorized image operations' the description mentions.
The sources do not state whether authentication is required beyond the ability to create a namespace, whether the flaw is reachable from the internet, or whether registries with tightly controlled namespace creation are affected. Do not assume either way; confirm with your vendor.

What to do now

Because no fixed version is named in the available sources, the first job is to find out what your vendor says about your specific build, then limit exposure while you wait or upgrade.

  1. Open the Red Hat CVE page and the Bugzilla entry and check whether your Quay version is listed as affected and whether a fix is available.
  2. Apply the vendor's fix or updated release as soon as one exists for your version, following your normal change process.
  3. As a precaution, restrict who can create namespaces on shared registries until you are patched, so untrusted users cannot create names containing an underscore.
  4. Rotate robot account tokens, starting with accounts that have write access to production repositories. Rotate after patching too, since the public record gives no way to rule out earlier exposure.
  5. Review which robot accounts exist, who owns them, and trim their permissions to the minimum repositories they need.
  6. Look through registry audit logs for unexpected pushes, pulls or tag changes made by robot accounts.
  7. Check recently created namespaces for names containing underscores that you cannot tie to a legitimate user.
Caution. Rotating robot tokens will break any pipeline or cluster that still uses the old value. Update the stored secrets in your CI/CD system and image pull configuration in the same maintenance window.

Detection and monitoring

Nothing in the sources describes indicators of compromise, so detection has to rely on general registry hygiene. Compare the images currently deployed with what your build system last produced, and investigate any tag that moved without a matching pipeline run. Alert on robot account activity outside normal hours or networks.

If your registry is reachable from the internet, keep an eye on its certificate and exposed hostname so an unexpected change stands out. A certificate and DNS change monitor can help with that, and the free SSL checker gives a quick one-off look. These only cover the outside of the registry and do not tell you whether this flaw is present. The wider habit of treating registry credentials like production secrets is covered in our small business security guide.

What is still unknown

  • Affected and fixed Quay versions are not given in the sources we have.
  • Whether the flaw is being exploited in the wild is not stated. The CVE data supplied to us shows no CISA KEV entry, but that does not confirm either way, so exploitation remains unconfirmed.
  • No proof-of-concept or technical write-up beyond the short NVD description is available to us.
  • Which Quay deployment types (self-managed or hosted) are in scope is not stated.

This article will need updating once a vendor advisory with version details is available. Until then, check your exposure with your vendor and apply the mitigation steps above, particularly if you run a multi-tenant registry.

Sources

This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.

  1. CVE-2026-110697: A flaw was found in Quay where the robot account listing function constructs a SQL LIKE pattern from user-controlled namespace names without escaping wildcard…nvd.nist.gov ·
  • Quay
  • CVE-2026-110697
  • container registry
  • SQL injection
  • credential theft