NGINX2Root: who found it, what it does, and how to check and fix your servers

Blue staircase of blocks with a single red block on the top step

On Wednesday 18 November 2026 the nginx team released versions 1.31.9 and 1.30.7 to fix a set of vulnerabilities known as NGINX2Root. nginx is the web server in front of a large share of the internet, including a lot of Laravel systems.

This post covers where the bugs came from, what an attacker can do with them, how to tell if you're exposed and how to fix it. Our own view on why this keeps happening is at the end.

Who found it, and how

Several teams found the main bug separately. nginx's release notes credit Xint, a security firm that builds AI-driven testing tools. Wiz says its own researchers, Rapid7, Metnir and Ikotas Labs reported the HTTP/2 bugs independently too.

Xint published the most detail and gave the set its name. They say one of their researchers found the main bug while preparing for ZeroDay Cloud 2026, and that a fully automated scan found four more a few days later with nobody guiding it. They found them on 11 October and reported them to F5, the company behind nginx, on 16 October. The details stayed private until the fix came out on 18 November.

Xint also found a separate bug in the Linux kernel that turns the nginx attack into full control of the machine. The NGINX2Root name covers all six.

What can happen

The worst of the five, CVE-2026-123007, lets somebody with no login run commands on your server. All they need is to reach a port where nginx is serving HTTP/2. The commands run with the same permissions as nginx.

From there the kernel bug takes them to root, which is full control of the server. Xint showed this working against servers on AWS, Google Cloud and Azure over the public internet. They also showed it breaking out of a Docker container to take over the host underneath.

F5's own description of the bug is more cautious. It says the bug crashes nginx, and that running code needs a standard memory protection called ASLR to be switched off or bypassed. Xint says its attack does bypass it, using nothing but the public HTTPS port. They have published how the bugs work and held back the last steps of the exploit. Rapid7, one of the other teams, is reported to have published proof-of-concept code. Treat it as remote code execution.

What it affects

  • • CVE-2026-123007 affects every official nginx release from 1.9.5 in 2015 up to 1.31.8, wherever HTTP/2 is switched on. That is common on HTTPS sites.

  • • The other four need a particular setup: HTTP/2 or HTTP/3 in front of a backend server, DNS lookups made while a request is being handled, or nginx's mail proxy.

  • • The kernel bug is in SCTP, a networking feature few web servers use. It has been in Linux since version 4.15 in 2018.

Xint counted 56.7 million hosts identifying as nginx on 10 November. Of the 11.94 million that reported a version number, 98.9% were in the affected range. That counts version numbers, so it shows the scale and doesn't prove each one can be attacked.

If you use a product that bundles nginx, check with whoever supplies it.

How to work out if you're vulnerable

Run these on the server.

1. Check the version

nginx -v

Fixed versions are 1.31.9 and above, and 1.30.7 and above within the 1.30 series. Anything from 1.9.5 to 1.31.8 has the bug unless your Linux distribution has patched it. More on that below.

2. Check whether HTTP/2 is switched on

sudo nginx -T 2>&1 | sed 's/#.*//' | grep -nE 'listen[[:space:]].*http2|^[[:space:]]*http2[[:space:]]+on[[:space:]]*;'

Any output means HTTP/2 is on. No output means it's off.

3. Check from outside

Run this from another machine, with your own domain in place of example.com.

openssl s_client -connect example.com:443 -servername example.com -alpn h2 </dev/null 2>&1 | grep -E 'ALPN protocol:|No ALPN negotiated'

ALPN protocol: h2 means the server accepts HTTP/2 from the internet. No ALPN negotiated means it doesn't. If a CDN such as Cloudflare sits in front of your site, this tests the CDN and not your own server.

A vulnerable version with HTTP/2 reachable from the internet needs fixing now.

If nginx came from your Linux distribution's own packages

The version number won't tell you. Ubuntu and Debian patch old versions without changing the number, so a fixed Ubuntu 24.04 server still reports 1.24.0. Check the full package version instead.

dpkg-query -W -f='${Version}  ${Maintainer}\n' nginx

Ubuntu published its fixed packages on 19 November (USN-9159-1). You're fixed at these versions or higher.

Ubuntu releaseFixed nginx package
26.101.30.4-7ubuntu1.2
26.04 LTS1.28.3-2ubuntu1.13
24.04 LTS1.24.0-2ubuntu7.20
22.04 LTS1.18.0-6ubuntu14.23
20.04 LTS1.18.0-0ubuntu1.7+esm5 (Ubuntu Pro only)
18.04 LTS1.14.0-0ubuntu1.11+esm6 (Ubuntu Pro only)

On Debian, check Debian's security tracker for the fixed version.

How to fix it

Update nginx, then restart it

On Ubuntu or Debian, whether nginx came from nginx.org or from the distribution:

sudo apt-get update
sudo apt-get install --only-upgrade nginx
sudo nginx -t
sudo systemctl restart nginx
nginx -v
  • • If the update asks whether to replace your config files, keep your existing ones.

  • • Restart nginx, don't just reload it. A reload re-reads the config and leaves the old program running.

  • • If you use Ubuntu's own package with automatic security updates switched on, the fix should install itself. Check it has, and restart nginx to be sure.

  • • If you installed nginx from nginx.org or another outside source, automatic security updates won't touch it. Somebody has to run the update.

If you can't update yet, switch HTTP/2 off

On nginx 1.25.1 and later, add http2 off; to the server block and remove http2 from every listen line. Both steps matter. We tested it, and http2 off; does nothing while a listen line still says http2. On older versions, http2 off; isn't recognised, so just remove http2 from the listen lines. Do this for every site on the same port. Then:

sudo nginx -t
sudo systemctl restart nginx

Run the outside check again to confirm it worked. Sites carry on working over HTTP/1.1. Anything that needs HTTP/2, such as gRPC, will stop. This blocks the main bug and one of the others. It does nothing for the remaining three, so it's a stopgap.

The kernel bug

It had no fix when it was disclosed. If your server doesn't use SCTP, stop it loading:

echo 'install sctp /bin/false' | sudo tee /etc/modprobe.d/disable-sctp.conf

This won't unload SCTP if it's already loaded (lsmod | grep sctp shows whether it is), and it won't help where SCTP is built into the kernel. Keep security updates switched on and reboot once the kernel fix arrives.

If your operating system is out of standard support

Your options are limited. nginx.org only builds packages for Ubuntu 22.04, 24.04 and 26.04, and Debian 12 and 13. Ubuntu 20.04 and 18.04 only get the fix through an Ubuntu Pro subscription. If neither applies, switch HTTP/2 off now and plan a move to a supported release.

Has it already been used against you?

An attack may leave almost nothing in the access log. Look for nginx workers crashing, and for nginx starting programs it has no reason to start.

sudo grep -Ei 'worker process.*exited|signal 11|segfault' /var/log/nginx/error.log

Crashes have ordinary causes too. Xint's write-up has a fuller checklist.

If somebody else hosts your system

Ask them two things. Has nginx been updated to a fixed version? Is the server on an operating system that still gets security updates? You should get a straight answer to both.

Our two cents

nginx published two security advisories in the whole of 2025. This year it has published 36, ten of them on Wednesday. PHP had its own security release on 21 October for two critical bugs, with another scheduled for 19 November.

The nginx bug had been there since 2015 and the kernel one since 2018. Nobody spotted either for years. Then five teams reported the HTTP/2 bugs separately before the fix came out, and Xint says an automated scan found four of its five with nobody guiding it. That's the change. Software now reads old code faster than people ever could, and attackers can use the same kind of tools.

So we expect this pace to continue. The useful question for a business is how quickly you can patch, and who does it. What we'd check:

  • • Know what you're running. You can't patch quickly if you first have to work out what you've got.

  • • Stay on supported versions. Old systems are the ones with the fewest options when something like this lands.

  • • Find out what "automatic updates" covers on your setup. It rarely covers everything.

  • • Use managed hosting where it fits, so the web server is the provider's job.

Plan for it, and make sure patching is somebody's job.

If you're not sure who patches your servers, get in touch and we'll take a look.

Sources

Things are moving quickly, so check these for anything newer than this post.

Codebased Limited. Registered in England and Wales, company number 14451447. Registered office: Rookery Offices, Rookery Farm, Wheaton Aston, Stafford, ST19 9QF.