What happened
On 18 November 2026, F5, the company behind the NGINX web server, released a fix for CVE-2026-121810. It is a use-after-free bug in the HTTP/3 support of both the free NGINX Open Source and the commercial NGINX Plus. Only servers with HTTP/3 enabled are affected.
The fix shipped together with several other security updates. In its advisory, the Slovak national CERT (SK-CERT) focuses mainly on a more serious HTTP/2 flaw, CVE-2026-123007, nicknamed “NGINX2ROOT”. It lists CVE-2026-121810 among four further flaws fixed by the same updates that may be exploitable in a similar way in specific configurations.
No exploitation of this flaw is known so far. It is not in the CISA KEV catalog, and the EU vulnerability database EUVD (entry EUVD-2026-128950) records no exploitation.
Why it is dangerous
Use-after-free means a program keeps using a piece of memory it has already released. Think of a cloakroom: you handed back your ticket, the hook went to someone else, yet you still reach for “your” coat using the old ticket. In software this causes a crash or, in the worse case, lets an attacker plant their own data in the freed spot.
According to F5’s description in the NVD, a remote attacker without any login can trigger the bug over the network in an NGINX worker process, which then restarts. Success also depends on conditions outside the attacker’s control. F5 says code execution is possible if ASLR (memory layout randomisation) is disabled on the server or if the attacker can bypass it. The flaw affects traffic handling only, not the management side of the system.
The most likely impact is therefore an outage or an unstable website. If code execution succeeded, it would run with the privileges of the nginx worker process. An attacker could then reach data the server handles and try to move further into the network.
- CVSS 8.1 out of 10 (high): the score was assigned by F5 using CVSS 3.1. It is pushed up by the attack working over the network without login, and pulled down by the high attack complexity. The nginx.org project rates the issue as “medium” in its own list.
- EPSS 1.75%: the model estimates the chance of exploitation within the next 30 days. That is low, but higher than roughly 78% of all other vulnerabilities (as of 19 November 2026).
- CISA KEV: the US agency CISA’s list of vulnerabilities known to be exploited. This flaw is not on it.
Who is affected
According to nginx.org, NGINX Open Source versions 1.25.0 to 1.31.8 are vulnerable. NGINX Plus with HTTP/3 enabled is affected too, with exact versions listed in F5 article K000163927. F5 did not evaluate versions that have reached end of technical support.
What matters is whether HTTP/3 is enabled. NGINX does not turn it on by itself, it has to be configured explicitly (the quic parameter of the listen directive). You will typically find it on modern websites, reverse proxies and load balancers in front of business applications, at web hosting providers and in containers built from newer NGINX images.
Servers without HTTP/3 are not affected by this particular flaw. You should still update them, because the same versions contain other bugs fixed in the same release, including CVE-2026-123007 in HTTP/2.
How to check if you are affected
The commands below only read information and change nothing. First check the version (for NGINX Plus the output also includes “nginx-plus”):
nginx -v
For containers, run the command inside the container:
docker exec <container-name> nginx -v
Then check whether HTTP/3 is enabled. If the command prints nothing, you are most likely not using HTTP/3:
sudo nginx -T 2>/dev/null | grep -E 'quic|http3'
From the outside, the alt-svc header helps. If it contains h3, the server advertises HTTP/3. Admins set this header manually, though, so its absence does not rule out HTTP/3:
curl -sI https://your-domain.com | grep -i alt-svc
Finally, check ASLR. A value of 2 means full protection:
cat /proc/sys/kernel/randomize_va_space
How to fix it
The fixed versions are NGINX Open Source 1.31.9 and 1.30.7 and anything newer. F5 lists the fixed NGINX Plus releases in article K000163927. The flaw is not in the CISA KEV catalog, so CISA has not set any required action for it.
- Start with internet-facing servers that have HTTP/3 enabled, then move on to internal systems.
- Update NGINX to 1.31.9 or 1.30.7 or later. SK-CERT warns that distributions such as Ubuntu and Debian may lag behind with packages and recommends the official nginx.org packages. If your distribution ships the fix without changing the version number, confirm it in its security notice.
- For containers, pull the fixed image and recreate the containers.
- Do a full restart of nginx, a reload is not enough. SK-CERT recommends this so that no old vulnerable processes keep running.
- Run
nginx -vagain to confirm the new version is live.
If you cannot update right away, temporarily disable HTTP/3 (remove quic from the listen directives and remove the http3 directive). The flaw only concerns HTTP/3, and browsers will automatically fall back to HTTP/2 or HTTP/1.1. F5’s description in the NVD does not list a specific workaround. Keep ASLR enabled, since according to F5 code execution is easier without it. Disabling HTTP/3 does not protect against the HTTP/2 flaw CVE-2026-123007, so updating is still essential.
What if you may already be compromised
Neither the vendor nor CERT teams have published specific indicators of compromise for this flaw. SK-CERT recommends watching above all for unexpected worker process crashes. The command also searches older compressed logs (adjust the path if your logs live elsewhere):
sudo zgrep -i 'exited on signal' /var/log/nginx/error.log*
- repeated worker crashes outside planned maintenance
- unknown processes running as the nginx or www-data user (
ps -eo user,pid,cmd | grep -E '^(nginx|www-data)') - new or modified files in web directories and in the nginx configuration
- new user accounts, SSH keys or scheduled tasks (cron)
Crashes alone do not mean a breach, the bug causes them even when an attempt fails. If you find unknown processes, files or accounts, disconnect the server from the network, do not delete anything and bring in an incident response specialist. Organisations in Slovakia can also report the incident to SK-CERT.