CVE-2026-18255: Quay Information Disclosure Vulnerability

CVE-2026-18255 Overview

A flaw was found in Red Hat Quay, the container image registry platform. Users configured in the GLOBAL_READONLY_SUPER_USERS list can view robot account tokens for repositories they are not members of. This authorization gap allows an attacker holding read-only superuser privileges to retrieve robot credentials and impersonate those robot accounts across the registry. The issue is tracked as CWE-863: Incorrect Authorization and carries a CVSS 3.1 base score of 7.2.

Critical Impact
Read-only superusers can extract robot account tokens for any repository and impersonate those service identities to push, pull, or modify container images.

Affected Products

Red Hat Quay (self-hosted container registry)
Deployments configured with GLOBAL_READONLY_SUPER_USERS
Downstream Red Hat products embedding the affected Quay component

Discovery Timeline

2026-07-29 - CVE-2026-18255 published to NVD
2026-07-30 - Last updated in NVD database

Technical Details for CVE-2026-18255

Vulnerability Analysis

The flaw resides in Quay's authorization checks around robot account token retrieval. Robot accounts in Quay function as non-human identities used by automation, CI/CD pipelines, and Kubernetes clusters to authenticate against repositories. Their tokens are equivalent to long-lived credentials for the repositories they scope to.

Quay supports a GLOBAL_READONLY_SUPER_USERS configuration for administrators who need registry-wide visibility without write privileges. The token-viewing endpoint fails to distinguish read-only superusers from repository members when deciding whether to return robot credentials. As a result, any account in that list receives cleartext robot tokens for repositories where it has no membership.

An attacker who obtains read-only superuser access, whether through credential theft, insider misuse, or misconfiguration, can enumerate robot tokens and reuse them. Reused tokens grant the robot's original permissions, which frequently include write access to production images.

Root Cause

The root cause is an authorization boundary error. The endpoint that surfaces robot tokens treats the read-only superuser role as sufficient for viewing secrets, rather than gating token disclosure on repository membership or write privileges. This is a classic CWE-863 incorrect authorization pattern.

Attack Vector

Exploitation requires network access to the Quay API and valid credentials for an account listed in GLOBAL_READONLY_SUPER_USERS. No user interaction is needed. Once authenticated, the attacker queries the robot-token endpoints for target repositories and captures returned tokens. The captured tokens are then presented to Quay or downstream registries to pull private images or push tampered images if the robot has write scope. See the Red Hat CVE-2026-18255 Advisory and Red Hat Bug Report #2508454 for vendor-confirmed technical details.

Detection Methods for CVE-2026-18255

Indicators of Compromise

Unexpected API calls to robot-token endpoints originating from accounts in GLOBAL_READONLY_SUPER_USERS.
Robot account authentications from source IPs or user agents that do not match the automation systems normally using those robots.
Image pushes or tag mutations performed by robot accounts outside their normal maintenance windows.

Detection Strategies

Audit Quay application logs for token retrieval operations correlated with the read-only superuser identity list.
Compare current membership of GLOBAL_READONLY_SUPER_USERS against change-management records to catch unauthorized additions.
Alert when a single account queries robot credentials across many unrelated repositories in a short window.

Monitoring Recommendations

Forward Quay API and audit logs to a centralized SIEM and build detections for robot-token read events by principal and repository scope.
Track robot account authentication sources and flag deviations from established baselines.
Review CI/CD registry activity for image pushes not tied to a corresponding pipeline run.

How to Mitigate CVE-2026-18255

Immediate Actions Required

Apply the Red Hat security update for Quay referenced in the Red Hat CVE-2026-18255 Advisory as soon as it is available for your deployment channel.
Review and minimize the GLOBAL_READONLY_SUPER_USERS list until patched, keeping only accounts with a documented operational need.
Rotate robot account tokens for any repository whose tokens may have been exposed to read-only superusers.

Patch Information

Refer to the Red Hat CVE-2026-18255 Advisory and Red Hat Bug Report #2508454 for fixed versions and errata mapping to Red Hat Quay releases. Apply vendor-supplied errata through your standard package or operator update process.

Workarounds

Remove all entries from GLOBAL_READONLY_SUPER_USERS until the patch is applied, and grant read-only visibility through narrower, per-namespace roles where possible.
Rotate robot tokens on a scheduled cadence and scope them to the minimum permissions each automation workflow requires.
Restrict network access to the Quay API to trusted administrative networks and CI/CD egress ranges.

bash

# Configuration example: audit and clear the read-only superuser list in config.yaml
# Before applying the patch, remove all entries from GLOBAL_READONLY_SUPER_USERS.
GLOBAL_READONLY_SUPER_USERS: []

# After removing entries, restart Quay so the configuration is reloaded.
# Then rotate any robot tokens that were exposed:
#   curl -X POST -H "Authorization: Bearer <admin-token>" \
#     https://<quay-host>/api/v1/organization/<org>/robots/<robot>/regenerate