On this page (7 sections)

A critical flaw tracked as CVE-2026-120984 affects the client management component of the Keycloak server. According to the GitHub Advisory Database entry, an authenticated user with client management permissions can bypass security controls, upload a malicious JavaScript policy, and have it executed on the server. The advisory says this could compromise the entire system.

If you run Keycloak as your login or single sign-on service, especially with delegated administrators or automated client provisioning, this is worth reviewing now. At the time of writing, the sources do not list affected or fixed versions, so the practical work is limiting who can manage clients and watching the vendor channels for a patch.

What happened

The advisory was published on November 7, 2026. It describes a flaw in how Keycloak handles a client that is created or imported with authorization services enabled. During setup, the server uses an internal session attribute and fails to clear it afterward. That leftover state lets a user with client management permissions get past a security control that should have stopped them.

The result is that the user can supply a JavaScript policy, and Keycloak runs it on the server when an authorization request is made. Code running inside the server process can reach whatever that process can reach, which is why the advisory talks about compromising the whole system.

Technical details in plain terms

The weakness class is CWE-94, code injection: the product builds or runs code using input that an outsider can influence, without properly restricting it. Keycloak's authorization services support policies written in JavaScript, which is a legitimate feature. The advisory says the stale session attribute is what allows the security controls around uploading such a policy to be bypassed.

The CVSS v3.1 vector in the advisory is AV:N/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:H. In plain terms:

  • Network attack vector and low complexity: it can be attempted remotely without special conditions.
  • High privileges required: the attacker must already hold client management permissions, so this is not an anonymous attack.
  • No user interaction needed.
  • Changed scope with high impact on confidentiality, integrity and availability: code execution on the server can affect things beyond Keycloak itself.
The 9.1 score is listed in the GitHub advisory, which is marked Unreviewed. Only one independent publisher covers this so far, and the NVD record repeats the same description rather than confirming it separately. Treat details beyond the description as unconfirmed.

Who is realistically exposed

Because high privileges are required, the biggest risk is not an outside stranger but anyone who can reach client management: a compromised administrator account, a stolen token for a service account with client management rights, or an over-trusted delegated admin, such as a team or tenant administrator in a shared realm.

  • Organizations that delegate client management to several teams or customers.
  • Setups that import or create clients automatically through scripts or CI pipelines.
  • Realms using authorization services, since the flaw concerns clients created or imported with them enabled.
  • Keycloak servers whose admin console or admin API is reachable from the internet.

Self-hosted deployments are the main concern. If you use a managed Keycloak offering, ask your provider whether they have assessed this CVE.

How to check your exposure

The sources do not name affected versions, so start by learning what you run and who can change clients. The Red Hat CVE page and Bugzilla entry are the places the advisory points to for further vendor information.

  1. Record the Keycloak version and distribution you run (upstream, or a vendor-packaged build).
  2. In the admin console, list every realm and note which ones have clients with authorization services enabled.
  3. List all users, groups and service accounts that hold client management roles in each realm, and remove any that are not needed.
  4. Review admin events and recent client creations or imports for anything unexpected, especially clients with authorization services and custom JavaScript policies you do not recognize.
  5. Check whether the admin console and admin API are exposed beyond your internal network.

What to do now

No patch or fixed version is named in the research, so the immediate goal is to shrink the group of people who could exploit this and to be ready to update quickly.

  1. Restrict client management permissions to the smallest set of trusted administrators and rotate their credentials.
  2. Require multi-factor authentication on all administrator accounts and review service accounts with client management roles.
  3. Limit network access to the admin console and admin API to a management network or VPN.
  4. Pause or review automated client import and creation pipelines until you know your exposure.
  5. Audit existing clients with authorization services and review any JavaScript-based policies for anything you did not create.
  6. Subscribe to the Keycloak project and your vendor's security notices, and check the Red Hat CVE page for fixed-version information.
  7. When a fixed release is published, test it in staging and apply it promptly, then restart the service during a planned window.
  8. Run Keycloak with a least-privilege operating system account and restricted outbound network access, so that code execution has less to work with.
Caution. Updating or restarting Keycloak interrupts sign-in for every application that relies on it. Schedule the change, back up your realm configuration and database first, and test in a staging environment.

Detection and monitoring

Without public exploit details, focus on behavior you can observe. Enable admin events in each realm and watch for client creation or import, changes to authorization settings, and new policies, particularly outside normal change windows or from unusual source addresses.

  • Alert on new or modified clients that have authorization services enabled.
  • Alert on logins to administrator accounts from new locations or at odd hours.
  • Watch the Keycloak host for unexpected child processes, new files or unusual outbound connections.
  • Keep admin and server logs somewhere an attacker with server access cannot easily erase.

If the admin console is reachable from outside, an outside-in view helps confirm what is exposed. Webforta Monitor can watch certificates, DNS and headers for changes on your login hostnames, and the free SSL checker gives a quick look at your certificate setup. Neither tells you whether this specific flaw is present.

If you suspect misuse, treat it as a server compromise: see our guide on what to do when a website is hacked, rotate client secrets and signing keys, and review every connected application.

What is still unknown

  • Which Keycloak versions are affected and which release fixes the issue.
  • Whether a workaround exists besides limiting permissions.
  • Whether the flaw is being exploited. The research shows no CISA KEV listing and reports no exploitation, but that is not proof that none is happening.
  • The advisory is Unreviewed on GitHub and the CVSS score there may be refined later.

We will not guess at these. Check the vendor references above for updates before relying on any version claim.

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-2026-120984 - GitHub Advisory Databasegithub.com
  2. CVE-2026-120984: A flaw was found in the Keycloak server's client management component.nvd.nist.gov ·
  • Keycloak
  • CVE-2026-120984
  • RCE
  • Identity management
  • CWE-94