NGINX Ingress Controller CVE-2026-55723

Exploitation requires the deployment to be configured using NGINX Ingress Controller CRDs or Ingress annotations (the affected configuration-generation paths), and the attacker must hold Kubernetes API permissions to create or modify those CRDs or Ingress annotations - consistent with CVSS PR:L. This is an authenticated attack (PR:L, not PR:N): an anonymous or read-only user cannot trigger it. The limiting factors are the required RBAC write access to Ingress/CRD objects and the fact that impact is confined to the control plane with no data plane exposure, so the realistic risk concentrates in multi-tenant clusters where ingress-write privileges are delegated to less-trusted tenants; in single-tenant clusters the required privilege is already held only by trusted operators.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:L) scores 8.7 High: network-reachable via the Kubernetes API, low complexity, but requiring low privileges (PR:L) - an attacker must already hold write access to Ingress objects, CRDs, or annotations. That privilege requirement is the key limiter: this is not an unauthenticated internet-facing bug but a privilege-escalation/tenant-isolation-breakout risk, most relevant in multi-tenant clusters where namespace admins should not be able to affect the shared ingress control plane. No EPSS score, KEV entry, or POC is present in the provided data, so there is no evidence of active exploitation - no public exploit identified at time of analysis. The vulnerability is a genuine priority for multi-tenant Kubernetes operators because it defeats RBAC boundaries, but for single-tenant clusters where Ingress writers are already trusted the practical risk is lower. The vendor explicitly states there is no data plane exposure, which caps availability impact (VA:L) and means the blast radius is the controller/control plane rather than proxied traffic.

In a multi-tenant cluster, a namespace administrator who is permitted to create Ingress resources or NGINX CRDs crafts an annotation or CRD field value that closes the intended directive context and appends attacker-chosen NGINX configuration. When the controller regenerates and reloads its configuration, the injected directives execute with the controller's privileges, allowing the attacker to create or delete files or disable ingress services. No public POC is identified; the low attack complexity and lack of required user interaction (AC:L/UI:N) make exploitation straightforward once the necessary write privilege is held.