NGINX Gateway Fabric CVE-2026-96615

Exploitation requires NGINX Gateway Fabric deployed with the Gateway Custom Resource Definition (CRD) AND an authenticated gRPC connection configured between the control plane and the NGINX Agent - this specific configuration is the precondition. The attacker must already hold valid credentials: a namespaced Kubernetes service account with access inside the cluster (PR:L, authenticated), limiting this to insiders or attackers who have already compromised a pod/workload. It is a control-plane-only issue with no data plane exposure, and exploitation is gated to cross-namespace key theft within a single cluster rather than remote unauthenticated access. Products past End of Technical Support are not evaluated, so version applicability may be limited.

This is a moderate, access-gated control-plane disclosure flaw rather than a mass-exploitation priority. The vendor CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:H) scoring 7.1 is internally consistent with the description: network-reachable gRPC, low complexity, but requiring low-level privileges (PR:L) - an attacker must already hold a valid service account inside the Kubernetes cluster. Impact is confidentiality-only (VC:H, VI:N/VA:N) - theft of TLS private keys - matching CWE-290 (Authentication Bypass by Spoofing / impersonation of the NGINX Agent gRPC connection). No EPSS score, CISA KEV listing, or public POC was provided in the input, so there is no public exploit identified at time of analysis and no evidence of active exploitation; the realistic risk is targeted abuse by an attacker who has already gained a foothold (a compromised pod/service account) in a shared multi-tenant cluster, not opportunistic internet-wide scanning. One nuance to verify: the description says impact is strictly control-plane with 'no data plane exposure,' yet stolen TLS private keys could enable downstream impersonation/MITM against the data plane later - the vendor scopes this as control-plane only (SC/SI/SA:N), so the immediate scored impact does not credit that secondary risk.

In a multi-tenant Kubernetes cluster running NGINX Gateway Fabric with the Gateway CRD and authenticated NGINX Agent gRPC connections, an attacker who controls a pod with a namespaced service account impersonates the authenticated Agent gRPC channel. By spoofing that trusted connection, they request and read the TLS private keys belonging to Gateway deployments in a different namespace, breaching tenant isolation. No public exploit is identified at time of analysis, so this would require custom tooling against the gRPC control-plane interface.