Source: CERT-EU
If you run NGINX anywhere in front of customers, this week deserves an hour of your engineering time. On 18 November 2026, CERT-EU published Security Advisory 2026-022, covering five vulnerabilities that F5 disclosed the same day. All five can let an unauthenticated remote attacker execute code with the privileges of the NGINX worker process, and researchers have already demonstrated exploitation of one of them in practice.
What was disclosed
The most urgent of the set is CVE-2026-123007 (CVSS 8.1), an integer overflow in HTTP/2 flow control that triggers an out-of-bounds write on the heap. The only precondition is HTTP/2 enabled on a listening port, which describes a very large share of modern web front ends. The remaining four issues are use-after-free flaws: CVE-2026-123633 (connection draining while proxying to an upstream over HTTP/2), CVE-2026-121810 (QUIC connection draining, so HTTP/3 listeners are in scope), CVE-2026-123634 (DNS resolver waiting list, and only reachable with a fairly specific resolver, proxy_pass and mirror configuration), and CVE-2026-113425 (mail proxy authentication lookups via auth_http).
Who is affected
The affected range is wide: NGINX Open Source mainline from 1.9.5 through 1.31.8 depending on the flaw, the stable 1.30.x branch including the current stable release 1.30.6, and NGINX Plus from 37.0.0 through 37.2.0.1 plus the R34, R35 and R36 releases. F5’s advisory lists the fixed versions. CERT-EU explicitly recommends updating as soon as possible and prioritising internet-facing instances, and notes that no exploitation in the wild has been reported yet. That last detail is the whole point of acting now: the window between disclosure and bulk scanning is usually measured in days.
Practical steps
Update your NGINX instances and base images first, internet-facing listeners before internal ones. If you cannot patch immediately, disabling HTTP/2 on exposed listeners (removing the http2 parameter from listen directives or setting http2 off;) removes exposure to the two HTTP/2 flaws. For monitoring, worker exits logged as “worker process … exited on signal” in the error log are a low-fidelity signal of attempts rather than proof of successful code execution, so pair the patch with proper process-creation logging if you need to justify a compromise assessment later.
For teams running NGINX inside containers, Kubernetes ingress controllers or VM golden images, this is also a good moment to check whether you can actually roll out a patched image fleet-wide within a week. If the honest answer is no, your delivery pipeline, not your web server, is the risk that needs attention first.
If you want a hand inventorying your NGINX footprint, planning the upgrade path or hardening your ingress configuration, contact us at Excello Digital and we will help you close this window before someone else opens it.
