Kubewarden: RBAC Reconnaissance via unchecked can_i host capability call

Kubewarden is a policy engine for Kubernetes. Prior to , An attacker with privileged AdmissionPolicy or AdmissionPolicyGroup create permissions (which isn't the default) can craft a policy that makes use of the can_i host callback. The callback issues a SubjectAccessReview (SAR) requests to enumerate RBAC permissions of any user or service account across the cluster. can_i does not perform that check to enforce the context-aware allow-list and forwards the request directly to the callback handler, which executes a real SubjectAccessReview using policy-server privileges. This creates a policy-level authorization gap: can_i is effectively usable even when the policy has no context-aware resource grant. This is an information disclosure / reconnaissance issue, and not direct workload data exfiltration. The attacker learns permission information, such as whether specific service accounts can "get secrets", "create pods", or "bind clusterroles" in chosen namespaces. This vulnerability is fixed in .

Problem type

CWE-862: Missing Authorization

Affected products

kubewarden

kubewarden-controller

< 1.35.0 - AFFECTED

References

https://github.com/kubewarden/adm-controller/security/advisories/GHSA-wqcw-g35j-j578

https://github.com/kubewarden/adm-controller/security/advisories/GHSA-wqcw-g35j-j578

x_refsource_CONFIRM

GitHub Security Advisories

GHSA-wqcw-g35j-j578

Kubewarden vulnerable to RBAC Reconnaissance via unchecked can_i host capability call

https://github.com/advisories/GHSA-wqcw-g35j-j578

Impact

Kubewarden is a policy engine for Kubernetes. Kubewarden cluster operators can grant permissions to users to deploy namespaced AdmissionPolicies and AdmissionPolicyGroups in their Namespaces. One of Kubewarden promises is that configured users can deploy namespaced policies in a safe manner, without privilege escalation.

An attacker with privileged AdmissionPolicy or AdmissionPolicyGroup create permissions (which isn't the default) can craft a policy that makes use of the can_i host callback. The callback issues a SubjectAccessReview (SAR) requests to enumerate RBAC permissions of any user or service account across the cluster. Three operations on the host capabilities kubewarden/kubernetes binding enforce the context-aware allow-list via can_access_kubernetes_resource():

list_resources_by_namespace
list_resources_all
get_resource

However, can_i does not perform that check and forwards the request directly to the callback handler, which executes a real SubjectAccessReview using policy-server privileges. This creates a policy-level authorization gap: can_i is effectively usable even when the policy has no context-aware resource grant.

This is an information disclosure / reconnaissance issue, and not direct workload data exfiltration. The attacker learns permission information, such as whether specific service accounts can "get secrets", "create pods", or "bind clusterroles" in chosen namespaces.

Patches

Cluster Operators, if providing their users with privileges to deploy AdmissionPolicies or AdmissionPolicygroups (which isn't the default), must then also deploy PolicyServers with reduced permissions for host capability calls. This includes the PolicyServer default. For that, make use of the new feature in v1.35:

For custom PolicyServers: Set the new PolicyServer.spec.namespacedPoliciesCapabilities , for example to an empty list [] which doesn't allow any capability.
For the default PolicyServer, set the .Values.policyServer.namespacedPoliciesCapabilities , for example to an empty list [] which doesn't allow any capability.

Also, if needed, they must ensure that those namespaced AdmissionPolicies or AdmissionPolicygroups are scheduled in the PolicyServers with reduced permissions. For that, they could make use of the new ns-policyserver-mapper policy, their own policy or other means, such as GitOps.

See: https://docs.kubewarden.io/howtos/policy-servers/namespaced-policies-capabilities

Workarounds

Cluster Operators can opt for:

Not allowing users to create namespaced policies (AdmissionPolicies, AdmissionPolicyGroups).
Removing SubjectAccessReview "create" permissions for the PolicyServer ServiceAccount RBAC being used, in custom PolicyServers and the PolicyServer default.

Resources

Code changes, with new security feature: https://github.com/kubewarden/kubewarden-controller/pull/1693
Documentation changes: https://github.com/kubewarden/docs/pull/737
- Explained new feature on 1.35.0
- Updated Threat model assessment

github.com

https://github.com/kubewarden/kubewarden-controller/security/advisories/GHSA-wqcw-g35j-j578

github.com

https://github.com/kubewarden/docs/pull/737

github.com

https://github.com/kubewarden/kubewarden-controller/pull/1693

docs.kubewarden.io

https://docs.kubewarden.io/howtos/policy-servers/namespaced-policies-capabilities

github.com

https://github.com/kubewarden/adm-controller/security/advisories/GHSA-wqcw-g35j-j578

nvd.nist.gov

https://nvd.nist.gov/vuln/detail/CVE-2026-42541

github.com

https://github.com/advisories/GHSA-wqcw-g35j-j578

JSON source

https://cveawg.mitre.org/api/cve/CVE-2026-42541

Click to expand

{
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2",
  "cveMetadata": {
    "cveId": "CVE-2026-42541",
    "assignerOrgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
    "assignerShortName": "GitHub_M",
    "dateUpdated": "2026-05-12T17:57:54.665Z",
    "dateReserved": "2026-04-28T16:56:50.190Z",
    "datePublished": "2026-05-12T17:57:54.665Z",
    "state": "PUBLISHED"
  },
  "containers": {
    "cna": {
      "providerMetadata": {
        "orgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
        "shortName": "GitHub_M",
        "dateUpdated": "2026-05-12T17:57:54.665Z"
      },
      "title": "Kubewarden: RBAC Reconnaissance via unchecked can_i host capability call",
      "descriptions": [
        {
          "lang": "en",
          "value": "Kubewarden is a policy engine for Kubernetes. Prior to , An attacker with privileged AdmissionPolicy or AdmissionPolicyGroup create permissions (which isn't the default) can craft a policy that makes use of the can_i host callback. The callback issues a SubjectAccessReview (SAR) requests to enumerate RBAC permissions of any user or service account across the cluster. can_i does not perform that check to enforce the context-aware allow-list and forwards the request directly to the callback handler, which executes a real SubjectAccessReview using policy-server privileges. This creates a policy-level authorization gap: can_i is effectively usable even when the policy has no context-aware resource grant. This is an information disclosure / reconnaissance issue, and not direct workload data exfiltration. The attacker learns permission information, such as whether specific service accounts can \"get secrets\", \"create pods\", or \"bind clusterroles\" in chosen namespaces. This vulnerability is fixed in ."
        }
      ],
      "affected": [
        {
          "vendor": "kubewarden",
          "product": "kubewarden-controller",
          "versions": [
            {
              "version": "< 1.35.0",
              "status": "affected"
            }
          ]
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "lang": "en",
              "description": "CWE-862: Missing Authorization",
              "cweId": "CWE-862",
              "type": "CWE"
            }
          ]
        }
      ],
      "references": [
        {
          "url": "https://github.com/kubewarden/adm-controller/security/advisories/GHSA-wqcw-g35j-j578",
          "name": "https://github.com/kubewarden/adm-controller/security/advisories/GHSA-wqcw-g35j-j578",
          "tags": [
            "x_refsource_CONFIRM"
          ]
        }
      ],
      "metrics": [
        {
          "cvssV3_1": {
            "version": "3.1",
            "vectorString": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N",
            "attackVector": "NETWORK",
            "attackComplexity": "LOW",
            "privilegesRequired": "LOW",
            "userInteraction": "NONE",
            "scope": "UNCHANGED",
            "confidentialityImpact": "LOW",
            "integrityImpact": "NONE",
            "availabilityImpact": "NONE",
            "baseScore": 4.3,
            "baseSeverity": "MEDIUM"
          }
        }
      ]
    }
  }
}