Linux distributions and appliance vendors often fix a vulnerability without changing the upstream version string. For CVE-2026-113425, that behavior is helpful for support lifecycles but dangerous for naive inventory. A host can report an old NGINX version and contain the patch, or report a newer-looking version while the running service still uses an old custom build. The closure decision belongs to the complete deployment, not the version banner.
Vulnerability overview
CVE-2026-113425 is a use-after-free in the NGINX mail proxy module. F5 says a specially crafted mail session can crash a worker, corrupt memory, or potentially execute code when ASLR is disabled or bypassable. The vendor classifies it as CWE-416 and fixed the upstream code in NGINX Open Source 1.31.9 and 1.30.7.
The disclosure roundup also identifies NGINX Plus 37.3.0.1 and 37.0.8.1 as fixed builds. Downstream packages may have their own fix notes and version schemes. Those records need to be tied to the actual installed artifact.
Mechanism and exploitation conditions
The defect is in mail proxy session handling. Public summaries point to auth_http processing and a client disconnect as the trigger area, but the full packet-level sequence is not public. The required deployment condition remains a reachable mail proxy built from affected code.
Backports change the provenance question, not the attack condition. A fixed patch must be present in the source used to build the running binary. A package changelog alone does not prove that the container layer, manually copied binary, or service unit points to that package.
Impact
If the verification is wrong, the worker remains exposed to crashes and memory corruption. The organization may also create a false audit trail that makes later incident analysis harder. In regulated or customer-facing environments, a remediation claim should distinguish package update, image deployment, process restart, and configuration validation.
Remote code execution remains conditional on ASLR being disabled or bypassable. That fact can influence communication, but it does not lower the patch priority below the need to remove a remote memory-safety flaw from a trusted relay.
Affected products and scope
Check every source of NGINX in the environment: operating-system packages, NGINX repositories, NGINX Plus packages, static builds, container images, and vendor appliances. The affected configuration is a mail proxy. An HTTP-only service can still be part of the NGINX fleet, but it does not exercise this module through the described path.
Inventory should include the package version, build identifier, binary hash, file path, container image digest, service start time, loaded modules, and effective configuration. These fields let an operator separate a vulnerable artifact from a patched one even when both display a similar upstream version.
Exposure context
ZoomEye returned 316,291,466 product fingerprints for app="NGINX" and 0 for the CVE-specific filter. The selected product count is a discovery lead. It does not distinguish upstream from backported code, arbitrary custom builds, or mail-enabled services.
Use the count to identify exposed or forgotten NGINX assets, then resolve package provenance internally. Downloaded scanners that mark every old version vulnerable can create unnecessary work; scanners that trust only the version string can miss fixed backports and unpatched custom binaries.
Remediation and mitigations
Select a vendor-fixed package or build, install it, restart the service, and verify the process image. For containers, deploy a new image digest and confirm the running container uses it. For virtual machines, compare the active service executable with the updated package payload. For NGINX Plus, record the support channel and exact fixed release.
Until closure, restrict untrusted access to mail listeners, remove unused mail proxy blocks, and run the worker with limited privileges. Keep the evidence package with the ticket: source version, patch note, binary hash, image digest, process start evidence, and successful configuration test. This makes the remediation repeatable across the next audit.