What changed
One patch window covers several code paths
NGINX and F5 published a set of fixes on 18 November. The clearest starting point is CVE-2026-123007: an HTTP/2 flow-control error that can corrupt a worker's memory and usually cause a crash. F5 says remote code execution is possible if address randomization is disabled or bypassed. The same release addresses separate flaws in HTTP/2 connection draining, HTTP/3, resolver handling and the mail proxy. Their prerequisites are not interchangeable.
For NGINX Open Source, the project lists 1.31.9 and 1.30.7 as not vulnerable to the named issues. NGINX Plus and related F5 products have their own affected-version tables. CERT-EU says researchers demonstrated the HTTP/2 issue, but reported no exploitation in the wild at the time of its 18 November advisory.
A crash is a lead, not a verdict
Defender workflow
Check the configuration before the conclusion
- 1
Inventory the active build and listeners. Record the running NGINX Open Source or Plus version, exposed listeners and whether HTTP/2 or HTTP/3 is enabled. Include ingress controllers and packaged appliances that use NGINX.
- 2
Match each enabled feature to its advisory. Review resolver, mirror and mail-proxy configuration where present. Use the vendor table for the exact product branch; a CVE name alone does not establish that a particular service is reachable.
- 3
Upgrade and verify the worker. Apply the appropriate fixed release, reload or restart as directed, and confirm the running worker uses it. If patching is delayed, CERT-EU notes that disabling HTTP/2 on exposed listeners mitigates the two HTTP/2 issues, with an operational tradeoff.
Defender note
Follow the protocol your clients can actually reach
For an internet-facing reverse proxy, prioritize the protocol and module exposed to untrusted traffic. A mail proxy issue needs a different owner and log review from an HTTP/2 listener. Keep those decisions separate in the change ticket so the team can tell what was fixed and what still needs attention.