/

CVE-2026-44774

Severity:

CRITICAL RISK

(9.9)

Improper Access Control in Traefik Kubernetes Gateway REST

CVE-2026-44774

Modified

Vulnerability type

Category:

Improper Access Control

CWE:

CWE-284, CWE-15

Impact:

Unauthorized Configuration Modification and Traffic Re-Routing

Severity:

CRITICAL RISK

(9.9)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Traefik's Kubernetes Gateway API provider can allow tenants with only HTTPRoute creation permissions to expose the internal REST provider handler, even when providers.rest.insecure=false is configured.

By abusing backend references ending with @internal, a low-privileged actor may gain live write access to Traefik's dynamic configuration, enabling unauthorized reconfiguration of routers and services in shared Gateway deployments.

Affected context

This vulnerability affects Traefik instances deployed with the Kubernetes Gateway API provider in multi-tenant or shared cluster environments.

It becomes relevant when:

  • Tenants can create or modify HTTPRoute resources.
  • The Traefik REST provider is enabled, but intended to be protected by providers.rest.insecure=false.
  • Internal services such as api@internal are exposed intentionally, while rest@internal is meant to remain inaccessible.

In such environments, the Gateway provider's handling of TraefikService backend references ending in @internal can unintentionally expose the REST handler.

Exploit preconditions

  • Attack surface: Kubernetes cluster running Traefik with the Gateway API provider enabled.
  • Required access: Ability to create or modify HTTPRoute resources within a namespace served by the shared Traefik Gateway (no direct Traefik admin access required).
  • Trigger point: Definition of a route whose backend reference is a TraefikService whose name ends with @internal, specifically targeting rest@internal.
  • Why exploitation is possible: The Gateway provider accepts arbitrary @internal backend names and does not distinguish between the intended api@internal and the sensitive rest@internal, allowing tenants to expose the REST provider despite providers.rest.insecure=false.

Why this happens

The underlying issue is an improper access control design (CWE-284) in how Traefik's Kubernetes Gateway API provider validates internal service references.

The provider treats any TraefikService backend whose name ends in @internal as valid, with insufficient checks to ensure that only intended internal endpoints (e.g., api@internal) are reachable from tenant-defined routes.

As a result, the security boundary implied by providers.rest.insecure=false is bypassed: the REST management interface (rest@internal), which should require stronger protections or be disabled, becomes routable via tenant-controlled HTTPRoute objects.

Conceptually, this reflects an authorization flaw where internal administrative surfaces are exposed through a shared data plane without proper partitioning or allowlisting.

Potential impact

  • An attacker with HTTPRoute creation permissions can gain write access to Traefik's dynamic configuration via the REST API.
  • Unauthorized creation, modification, or deletion of routers, middlewares, and services.
  • Traffic hijacking or rerouting to attacker-controlled backends, enabling data interception or manipulation.
  • Downtime or denial of service by misconfiguring routes or removing critical services.
  • Potential lateral movement or escalation within the cluster by exposing additional internal services or manipulating ingress paths.

General mitigation guidance

Effective mitigations focus on tightening access control boundaries and reducing exposure of administrative endpoints:

  • Restrict which internal services can be referenced from tenant-controlled Gateway resources using allowlists, explicit kind checks, or dedicated service names that cannot collide with admin endpoints.
  • Treat all management and configuration APIs (like the REST provider) as sensitive admin surfaces and segregate them from tenant traffic, for example via separate entrypoints, networks, or instances.
  • Apply least-privilege principles to Kubernetes RBAC so that only trusted operators can define routes that may indirectly surface internal Traefik services.
  • Use multi-tenant isolation strategies (separate Traefik instances or dedicated Gateways) to avoid sharing administrative control paths across untrusted tenants.
  • Regularly review configuration for internal handlers and remove or harden any that are not strictly required in the current deployment model.

How teams detect issues like this

Detection typically combines configuration inspection, logging review, and policy enforcement:

  • Implement admission controllers or policy engines (e.g., OPA/Gatekeeper, Kyverno) that reject HTTPRoute or TraefikService objects referencing *@internal backends except from explicitly trusted namespaces or service names.
  • Use static or configuration analysis tools to scan cluster manifests and Traefik configs for routes targeting internal admin surfaces such as rest@internal.
  • Monitor Traefik access logs for unexpected access to REST handler paths or unusual configuration changes originating from non-admin tenants.
  • During code and configuration review, flag patterns where internal control-plane endpoints are reachable through shared data-plane routing rules.
  • Periodically audit dynamic configuration state against a known-good baseline to detect unauthorized routers, services, or middlewares.

Sources

SourceLink
github.comgithub.com/traefik/traefik/releases/tag/v2.11.46
github.comgithub.com/traefik/traefik/releases/tag/v3.6.17
github.comgithub.com/traefik/traefik/releases/tag/v3.7.1
github.comgithub.com/traefik/traefik/security/advisories/GHSA-96qj-4jj5-wcjc
access.redhat.comaccess.redhat.com/errata/RHSA-2026:36820
access.redhat.comaccess.redhat.com/security/cve/CVE-2026-44774
bugzilla.redhat.combugzilla.redhat.com/show_bug.cgi?id=2477937
security.access.redhat.comsecurity.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-44774.json

Frequently asked

questions

It can be, depending on how Traefik is exposed. If the shared Gateway entrypoint is publicly reachable and tenants with HTTPRoute permissions can define routes, an external attacker who controls such a tenant could remotely reach the REST handler.

The attacker needs the ability to create or modify HTTPRoute (and associated backend references) in a namespace served by the affected Traefik Gateway. Direct Traefik admin or cluster-admin permissions are not required.

Ensure that administrative interfaces like the REST provider cannot be exposed via tenant-controlled routing, ideally by isolating admin surfaces, strictly validating internal service references, and enforcing that only a minimal, allowlisted set of internal endpoints is routable.

When integrating control-plane features with multi-tenant data planes, design explicit authorization boundaries: treat internal/admin services as separate trust zones, validate all cross-zone references against strict policies, and apply least-privilege RBAC so tenant-level controls can never directly surface or manipulate administrative interfaces.

Find out what's already
exploitable in your codebase.