security

NGINX HTTP/2 flaws (CVE-2026-123633 and CVE-2026-123007): what we're doing and what you should do today

Two NGINX flaws can let an unauthenticated attacker run code on any server with HTTP/2 enabled. Here's our exposure, what we're doing for clients, and the steps we recommend right now.

Alluxi·November 20, 2026·7 min read

Published November 20, 2026. We'll update this post as the situation develops.

On November 18, the NGINX project released versions 1.30.7 and 1.31.9, which fix ten vulnerabilities. Two of them, CVE-2026-123633 and CVE-2026-123007, are in the HTTP/2 code and can let an unauthenticated attacker run code on the server. F5, which maintains NGINX, scores both 9.2 (Critical) under CVSS v4.0. Hosting providers, including DigitalOcean, have started emailing customers about them.

The flaws are in NGINX itself, not in the website or application behind it. A server can be vulnerable even when the application code is fine. Wiz, one of the teams that reported the flaws, demonstrated code execution against a server that only served static files, with no backend application and no third-party modules.

The short version

  • Affected: NGINX Open Source 1.9.5 through 1.31.8 with HTTP/2 enabled. CVE-2026-123007 covers that whole range. CVE-2026-123633 starts at 1.19.7. NGINX Plus R33 and the 37.x generation are affected too.
  • Fixed in: NGINX 1.30.7 (stable) and 1.31.9 (mainline). For NGINX Plus, 37.0.8.1 or 37.3.0.1 include both fixes.
  • If you can't patch today: turn HTTP/2 off. That blocks both flaws until you upgrade.
  • Exploitation: as of this writing, nobody has reported attacks in the wild. Wiz and Xint are holding back their working exploits. Rapid7 has reportedly published proof-of-concept code for CVE-2026-123007, so we expect that to change.

What the two flaws are

CVE-2026-123007 is an integer underflow in HTTP/2 flow control. A remote client can push NGINX's send-window accounting out of range, which corrupts memory. It has been in the code since HTTP/2 support landed in 1.9.5, eleven years ago.

CVE-2026-123633 is a use-after-free in HTTP/2 connection handling. When a worker runs low on free connections, NGINX can close a connection and free memory that an active request still uses. Reverse proxies under load are the most exposed.

Turning either bug into code execution takes work. F5 says it is possible when ASLR is disabled or the attacker can bypass it, and Wiz says it bypassed ASLR on official builds. The reported proof of concept needs to send about 128 GB of HTTP/2 data to one server. That is a real barrier, but not one we'd plan around. The NGINX project itself rates the flaws "major" and "medium" on its own scale, lower than the CVSS score suggests. We're treating them as critical.

Is Alluxi affected?

alluxi.com and our internal tools are Next.js applications on Vercel. We don't run NGINX servers for them, and neither the website nor its forms are exposed through an NGINX instance we manage.

Some of the systems we build and maintain for clients do use NGINX, usually as the reverse proxy in front of a Node.js, Django or Laravel application, or inside a container image. Those are where our team is focused.

What we're doing

  1. Inventory. We're listing every client system we built or maintain where NGINX runs. That covers production and staging servers, load balancers, container base images, Kubernetes ingress controllers and marketplace images that ship with NGINX preinstalled.
  2. Patching. We're moving to 1.30.7 or 1.31.9, or to the patched package from the server's Linux distribution, and rebuilding container images from patched base images.
  3. Restarting, not reloading. After a binary upgrade, a reload leaves the old master process running the old code. We restart NGINX and confirm the running process is the new build.
  4. Turning HTTP/2 off where we can't patch yet. On servers that need a maintenance window or have no distribution fix yet, we disable HTTP/2 until the upgrade is done.
  5. Checking for compromise. We're reviewing logs on affected servers for the warning signs listed below.
  6. Contacting you directly. If we maintain a system for you that uses NGINX, your project lead will contact you with its status and any decision we need from you. You don't need to open a ticket.

What we recommend, whether or not you work with us

1. Find every place NGINX runs. NGINX is often there without anyone having chosen it: the reverse proxy in a server setup tutorial, a 1-Click or marketplace image, a Docker Compose file, a Kubernetes ingress controller, a control panel, or a vendor appliance. Check staging servers and old subdomains too. Then check which of them have HTTP/2 on:

Need help with security?

Our team has 10+ years building custom solutions. Tell us about your project and we'll reply the same business day.

Learn about our cloud services
sudo grep -rn http2 /etc/nginx/
code

2. Patch, and check the right version number. If you installed NGINX from nginx.org, upgrade to 1.30.7 or 1.31.9 and confirm with nginx -v. If you installed it from your Linux distribution, the version number alone won't tell you. Ubuntu, Debian and others backport the fix and keep the old version number, so compare your package version (dpkg -l nginx or rpm -q nginx) against your distribution's advisory. As of November 20, Ubuntu, Debian 12 and 13, RHEL 10, Amazon Linux and Alpine have shipped fixes. RHEL 8 and 9 and Fedora have not.

3. Restart NGINX. A reload is not enough after a binary upgrade. Restart the service so the patched binary is the one serving traffic. For containers, pin the image tag (nginx:1.30.7 or nginx:1.31.9), rebuild, redeploy, and confirm the digest of the image that is running.

4. If you can't patch today, turn HTTP/2 off. On NGINX 1.25.1 or later, set http2 off; in each affected server block (or once in the http block) and remove the http2 parameter from every listen line. On older versions the listen parameter is the only switch, and one server block that keeps it turns HTTP/2 on for every site sharing that IP and port. Then test and reload:

sudo nginx -t && sudo systemctl reload nginx
curl -so /dev/null -w '%{http_version}\n' --http2 "keyword">https://example.com/
CSS

The second command should print 1.1. Browsers fall back to HTTP/1.1 without visitors noticing much. Anything that requires HTTP/2, such as native gRPC, will break, so test before you do it in production. Confirm the old worker processes have exited, and turn HTTP/2 back on only after upgrading.

5. Don't count on your CDN. A CDN that speaks HTTP/1.1 to your origin keeps these requests away from it, but only if attackers can't reach the origin directly. If your server's IP still answers on port 443, the CDN doesn't protect you. Restrict the origin to your CDN's IP ranges.

6. Look for warning signs. Nobody has published confirmed indicators of compromise yet. These are the general signs researchers suggest watching for:

  • NGINX worker processes crashing. Search the error log for exited on signal.
  • An NGINX worker starting a shell or any other program.
  • Long-lived HTTP/2 connections that receive large amounts of data while logging few or small requests.
  • Unusual outbound connections or DNS activity from the server.

A crash on its own doesn't prove a breach. A worker launching another program is a much stronger signal. If you see it, treat the server as compromised and rotate the credentials stored on it.

7. Remember the other eight fixes. The same release fixes eight more vulnerabilities, in HTTP/3, the DNS resolver, the mail proxy, TLS 1.3 request handling and other modules. Turning HTTP/2 off does nothing for those. It buys you time, and the upgrade is still the fix.

8. Ask your vendors. Products that bundle NGINX need a fixed build from their own vendor. That includes ingress controllers, API gateways, proxy images and forks such as OpenResty. If you still run the community ingress-nginx controller for Kubernetes, note that the project was retired in March 2026 and no longer ships fixes.

Need a hand?

If you're not sure whether you're exposed, or you have servers that nobody on your team has touched in a while, contact our team. We'll help you work out where NGINX is running and what to do first.

Sources

Share this post

Related posts

Related services

Explore how we can help with your next project

Schedule a free technical consultation

Tell us briefly about your project and a senior team member will reply the same business day.

alluxi logo

Get tips, case studies, and stay up to date on Alluxi.