All articles

NewsNovember 12, 20264 min read

Red Hat Quay: critical flaw exposes and alters private images

CVE-2026-123710 (CVSS 9.8) in Red Hat Quay: anyone who reaches the key server can sign tokens and read or overwrite every image in the registry.

Hands holding a cloud with a padlock, surrounded by data and connection icons

Red Hat has published an advisory for a critical vulnerability in Red Hat Quay, the container image registry companies use to store, manage and distribute the images of their applications. The flaw, CVE-2026-123710 with a CVSS score of 9.8, makes it possible to take over the signing of registry tokens and therefore to read and overwrite the images of any repository, private ones included.

The details below come from Red Hat’s advisory and Bugzilla report, as reported by SecurityWeek. The Hacker News also covered the story with the same picture: reading and overwriting all private images in a registry.

What happened

According to Red Hat, the problem lies in Quay’s service key server, the service that manages the keys used to sign registry tokens. Affected are deployments where this service is reachable by untrusted clients. Red Hat describes the flaw as a takeover of token signing by an unauthenticated attacker.

The vulnerability was discovered by Red Hat’s Dave O’Connor. At the time of reporting, neither the advisory nor the Bugzilla report identified a fixed version.

How the attack works

The chain described by Red Hat is as follows:

  1. Initial token. If anonymous access is enabled, an attacker without an account obtains a registry token. If anonymous access is disabled, any valid Quay account is enough: this is why, Red Hat warns, turning off anonymous access does not protect you.
  2. Attacker key. By exploiting weaknesses in the key rotation mechanism, the attacker installs their own signing key, which the registry treats as trusted.
  3. Arbitrary tokens. With that key they generate tokens with read and write permissions on any repository, bypassing access controls.

The consequences are twofold. On one side, the attacker can download private images and everything they contain, including credentials and other secrets left in the image layers. On the other, they can overwrite images that other systems will later pull: build pipelines, clusters and production servers.

Red Hat points out three aspects that make the compromise persistent and hard to clean up:

  • the malicious key can be configured without an expiration date and survives registry restarts until it is removed;
  • rotation deletes the legitimate instance key, and the attacker key can delete replacement keys: the registry may stop working for legitimate users, who can no longer push or pull images;
  • changing passwords or regenerating robot tokens does not invalidate an unauthorized signing key.

In MITRE ATT&CK terms the chain involves Exploit Public-Facing Application (T1190), if the service is exposed, Forge Web Credentials (T1606) for the forged tokens, Implant Internal Image (T1525) for overwriting images and Unsecured Credentials (T1552) for the secrets stored in images.

Versions and fixes

According to SecurityWeek, fixes for this flaw and for another Quay vulnerability, CVE-2026-110697, were merged into the project’s GitHub repository on 11 November. The sources do not yet say which released versions of Red Hat Quay include the fix: follow Red Hat’s official advisory. Red Hat warns that, until the fix is installed, an attacker can reinstall malicious keys even after they have been removed.

Impact for European and Italian companies

An image registry is a central point of the software supply chain: what it holds ends up in production clusters and servers. A replaced image runs the attacker’s code wherever it is started, with the permissions of the service using it. Private images also often contain configurations, API keys and credentials for databases and cloud services.

For entities in scope of NIS2, vulnerability handling and supply chain security are among the required measures: a registry compromise can turn into a significant incident, with reporting obligations. The same applies to healthcare and critical services that deploy containerised applications.

What to do now

The first three measures are those recommended by Red Hat; the others are good practices to check whether the registry has been compromised.

  • Restrict access to the key service. Make it reachable only from trusted networks and systems, never from the internet or from uncontrolled clients.
  • Review logs for suspicious changes to service keys, and remove unauthorized keys, bearing in mind they could be reinstalled until the fix is in place.
  • Install the fixed version as soon as Red Hat lists it in the advisory.
  • Do not rely on disabling anonymous access, changing passwords or regenerating robot tokens: they are not enough.
  • Check image integrity: compare the digests of the images in the registry with those produced by your build pipelines and, if you sign your images, enforce signature verification before deployment. If in doubt, rebuild and republish from trusted builds.
  • Rotate the secrets contained in private images if you find unauthorized keys, and consider moving them to a secrets manager instead of keeping them in images.

How ARXDOME helps

ARX Expose and ARX Scout show which of your company’s services are exposed to the internet, so a registry reachable from outside does not go unnoticed. ARX Shield keeps track of open vulnerabilities on monitored machines, while ARX Sentry collects server logs and events for SOC analysis.

Last updated:

Any questions?

Let’s talk about your company’s needs, with no obligation.

Contact us

Related articles