If your NGINX has HTTP/2 enabled and runs anything older than 1.30.7 (stable) or 1.31.9 (mainline), patch it. If you cannot patch today, turn HTTP/2 off. Those are the fixed releases listed by nginx.org for CVE-2026-123007 and CVE-2026-123633, two HTTP/2 flaws disclosed on November 18, 2026 (per Wiz).[2][1] Wiz Research, which co-discovered them, says both need HTTP/2 enabled and rates both Critical (CVSS 4.0 score 9.2).[1] This guide is the triage version: a table to decide whether you are exposed, a table of fixed versions, the commands to check your server, the temporary mitigation, and what is and is not confirmed about the side effects.
Everything below comes from pages we opened on November 19, 2026: Wiz's write-up, the nginx.org advisory list, the nginx HTTP/2 module docs, nginx.org's process-control page, and two F5 advisories. We did not run NGINX for this post and make no testing claims; the commands are standard NGINX and curl usage that you should run in staging first.
What the two bugs are
Both are memory-safety bugs in how NGINX handles HTTP/2 traffic.[1]
| CVE-2026-123007 | CVE-2026-123633 |
|---|
| Bug | Integer underflow in HTTP/2 flow control (send_window) | Use-after-free on an HTTP/2 connection |
| Affected OSS versions | 1.9.5 to 1.31.8 | 1.19.7 to 1.31.8 |
| Extra condition | Wiz demonstrated code execution over HTTPS on 64-bit systems serving ordinary static files, no backend or third-party module | Affects reverse-proxy processing when available connection slots run low (F5: worker is low on worker_connections) |
| F5 impact wording | Crash the worker, corrupt memory, or potentially execute code remotely | Denial of service or possible code execution |
Both F5 advisories add the same ASLR caveat: attackers can execute code on systems with ASLR disabled, or when the attacker can bypass it.[4][5] Wiz goes further and says it confirmed code execution with an ASLR bypass on official NGINX 1.31.7 and 1.31.8 builds for CVE-2026-123007.[1] Wiz also says it built proof-of-concept exploits and is not releasing them; none of the pages we read report exploitation in the wild.[1]
Am I exposed? A decision table
Start with the one condition Wiz states in bold: both vulnerabilities require HTTP/2 to be enabled.[1] Then work down the rows.
| Your setup | CVE-2026-123007 | CVE-2026-123633 |
|---|
| HTTP/2 not enabled on any listener | Precondition not met (Wiz: HTTP/2 required) | Precondition not met (Wiz: HTTP/2 required) |
| HTTP/2 on, serving static files directly over TLS | In scope (Wiz's demonstrated path) | Needs the proxy path, so not the demonstrated case |
| HTTP/2 on, reverse proxy to an app or API | In scope | In scope |
| HTTP/2 on, version 1.30.7 / 1.31.9 or newer | Fixed per nginx.org | Fixed per nginx.org |
| A CDN or WAF terminates HTTP/2 and talks to NGINX over something else | Wiz says this "may mitigate parts of the exploit" and "depends on implementation." Treat it as partial cover at best, not a patch. |
One detail worth acting on: a server on 1.9.5 through 1.19.6 is only in the range for CVE-2026-123007, while 1.19.7 and later is in both.[2] Wiz's own table says CVE-2026-123007 is "Exploitable in default configurations with HTTP/2 enabled" and CVE-2026-123633 is "Exploitable configured with HTTP/2 as a proxy server."[1] If you run NGINX as a proxy, assume both apply.
Step 1: check your version and whether HTTP/2 is on
The HTTP/2 module is not built by default; it needs the --with-http_v2_module configure parameter.[3] So check both the build and the live config.
# version, and whether this build includes the HTTP/2 module
nginx -v
nginx -V 2>&1 | tr ' ' '\n' | grep http_v2
# dump the full effective config and find HTTP/2 settings
sudo nginx -T 2>/dev/null | grep -nE 'http2|listen'
In the nginx -T output, any listen ... http2; line or http2 on; directive means HTTP/2 is enabled for that server. The http2 directive appeared in 1.25.1, defaults to off, and is valid in the http and server contexts, so a single line near the top of the config can cover every server block.[3] Scan for http2 on in an included file too; nginx -T prints those.
If your NGINX came from an operating system package, the version string may not match the upstream numbers in this post. The pages we read do not say how distributions version their patched builds, so check your vendor's advisory rather than comparing numbers alone.
Step 2: compare against the fixed versions
nginx.org lists both CVEs as "Not vulnerable: 1.31.9+, 1.30.7+."[2] Wiz's fix table adds NGINX Plus.[1] F5's advisories add other NGINX products.[4][5]
| Product | Fixed release |
|---|
| NGINX Open Source, stable branch | 1.30.7 |
| NGINX Open Source, mainline branch | 1.31.9 |
| NGINX Plus, fixes both CVEs | 37.0.8.1 or 37.3.0.1 (the 123633 fix alone shipped in 37.0.7.1 and 37.2.0.1) |
| NGINX Instance Manager | 2.25.0 (both advisories) |
| NGINX Gateway Fabric 2.x | 2.9.0 (both advisories) |
| NGINX Ingress Controller | LTS: 2026-lts-r11; 5.x: 5.8.0 (both advisories) |
That Plus row matters: the two CVEs landed in different Plus point releases, so 37.0.7.1 patches the use-after-free but not the flow-control bug. Take the later number.[1] We could not confirm Plus R33 to R36 fix versions from the pages we read, so Plus users on those branches should read F5's tables directly.[4]
Step 3: the temporary mitigation
Wiz's guidance if you cannot upgrade immediately: disable HTTP/2 temporarily. On NGINX 1.25.1 or later, set http2 off; in the affected server blocks and remove the http2 parameter from any listen directives. Then test application compatibility, reload, and confirm the old workers have exited. Re-enable HTTP/2 after upgrading.[1] On versions older than 1.25.1 the http2 directive does not exist (it appeared in 1.25.1), so the listen parameter is the only switch.[3]
# before (any version)
server {
listen 443 ssl http2;
server_name example.com;
}
# after, NGINX 1.25.1 or later
server {
listen 443 ssl;
http2 off;
server_name example.com;
}
# after, NGINX older than 1.25.1: just drop the parameter
server {
listen 443 ssl;
server_name example.com;
}
Here is the catch. Both F5 advisories list the mitigation as "None."[4][5] The workaround comes from Wiz, whose own research says both flaws need HTTP/2, not from F5's advisory text. It follows from the stated precondition, but the nginx.org and F5 pages do not endorse it. Treat http2 off as a stopgap that Wiz recommends, and upgrade as the actual fix.
Step 4: reload and confirm the old workers are gone
On reload, nginx starts new worker processes with the new configuration and asks the old ones to shut down gracefully; old workers keep serving existing clients and exit only after those are done.[6] Wiz tells you to confirm the old workers have exited, so do not assume the reload finished the job.[1]
sudo nginx -t && sudo nginx -s reload
# worker start times should all be after the reload
ps -eo pid,lstart,cmd | grep '[n]ginx: worker'
# from another machine: should print 2 if HTTP/2 negotiates, 1.1 if it does not
curl -sI -o /dev/null -w '%{http_version}\n' --http2 https://example.com
If a worker from before the reload is still listed, nginx's docs say it is still servicing old clients.[6] The pages we read do not say how long to wait or whether to force it; that call depends on your traffic. If you script the curl check, our cURL to code converter turns it into a snippet for your language, and the HTTP status codes reference helps when a health check starts returning something other than 200.
What breaks when you turn HTTP/2 off
Less is confirmed here than most guides imply. What the sources support:
Wiz says to test application compatibility before reloading. It does not list specific breakages.[1]
HTTP/2 over TLS is negotiated through the ALPN extension.[3] The module exposes a
$http2variable that ish2for HTTP/2 over TLS,h2cfor cleartext HTTP/2, and an empty string otherwise.[3]The module's HTTP/2 push directives (
http2_push,http2_push_preload) are marked obsolete since 1.25.1.[3]http2_max_concurrent_pushescarries the same obsolete-since-1.25.1 note, so a config that still sets push directives is carrying dead settings either way.[3]
That $http2 variable is the practical tool. Add it to a log format for a day before you flip the switch and you will see how much of your traffic is actually h2:
# both lines go in the http context
log_format h2check '$remote_addr "$request" $status http2=$http2';
access_log /var/log/nginx/h2check.log h2check;
What we could not confirm from any page we opened: the performance cost of falling back to HTTP/1.1 for your workload, whether specific clients or gRPC upstreams depend on HTTP/2 on the client side, and how the other http2_* tuning directives behave once HTTP/2 is off. Test those in staging. A mitigation that quietly breaks a gRPC client is a worse day than a planned upgrade.
Docker and Kubernetes
None of the five primary pages we read discusses container images, so we are not going to guess which image tags carry 1.30.7 or 1.31.9. What is verifiable: a container runs an NGINX binary, and the version check works the same inside it. Pull or rebuild, restart, then ask the running container, not your registry listing.
# Compose: pull the new image, recreate, then check the running binary
docker compose pull
docker compose up -d
docker compose exec web nginx -v
Replace web with your service name. If you build your own image from a pinned base, update the pin first or the rebuild will reproduce the old version. For the Compose side of this, see our docker run to Compose flag guide and the Compose production guide.
On Kubernetes, F5's advisories list NGINX Ingress Controller and NGINX Gateway Fabric as affected, with the fixed releases in the table above.[4] Those products embed NGINX components, so upgrading the controller is the fix, not editing a config in a pod.
nginx.org, F5 and Wiz rate these differently
You will see "9.2 Critical" in headlines. Here is exactly what each source says.
| Source | CVE-2026-123007 | CVE-2026-123633 |
|---|
| Wiz | Critical, CVSS 4.0 score 9.2 | Critical, CVSS 4.0 score 9.2 |
| F5 advisory | High 8.1 (CVSS v3.1); Critical 9.2 (CVSS v4.0) | High 8.1 (CVSS v3.1); Critical 9.2 (CVSS v4.0) |
| nginx.org advisory list | Severity: major | Severity: medium |
So the 9.2 is the CVSS 4.0 figure, and F5 publishes it alongside an 8.1 under CVSS 3.1 for both bugs.[4][5] The nginx.org list uses its own severity words and prints no CVSS number; it calls the flow-control bug "major" and the use-after-free "medium."[2] They are different scales, so the labels do not line up one to one. The practical reading: do not downgrade CVE-2026-123633 because nginx.org says "medium" if you run a proxy. F5's own description of it includes possible code execution, and Wiz says it is more likely to be exploited in a proxy configuration.[1]
The same release fixes other CVEs
The November 18 release is not just these two. nginx.org lists a resolver use-after-free (CVE-2026-123634), an HTTP/3 use-after-free (CVE-2026-121810), and several others as fixed in 1.31.9 and 1.30.7, so upgrading closes more than the HTTP/2 pair.[2] Wiz's table lists the HTTP/3 and mail-proxy issues as separate configuration conditions. Wiz's HTTP/2 mitigation does not address those. If you run HTTP/3 or the mail proxy module, the sources we read offer the upgrade, not a config switch.[1]
A 15-minute triage order
Run
nginx -vand thenginx -T | grepcommand on every edge host. Anything with HTTP/2 off is out of scope for these two bugs.Rank what remains by reachability: internet-facing proxies first, since Wiz says exploitation assumes the attacker can reach the server directly.[1]
Upgrade to 1.30.7 or 1.31.9 where you can. Where you cannot yet, apply
http2 off;and reload.Confirm the old workers exited, then re-check with curl.
Watch for unexpected worker crashes or commands launched by NGINX. Wiz calls these general warning signs, not confirmed indicators of these vulnerabilities.[1]
FAQ
Do I need to patch if HTTP/2 is off?
For these two CVEs, no: Wiz states both require HTTP/2.[1] The same release fixes other issues (HTTP/3, mail proxy, resolver) that have their own conditions, so you will want the upgrade anyway.[2]
Is a version below 1.30.7 always vulnerable?
With HTTP/2 on, a version inside the listed ranges is affected: 1.9.5 to 1.31.8 for CVE-2026-123007 and 1.19.7 to 1.31.8 for CVE-2026-123633, which also needs the reverse-proxy condition.[2][4][5] Packaged builds from a distribution may differ; check your vendor.
Does a CDN in front of NGINX protect me?
Possibly in part. Wiz says CDNs and WAFs that terminate HTTP/2 may mitigate parts of the exploit and that it depends on implementation.[1] It is not a substitute for the patch.
Can I fix this with a config change alone?
F5 lists the mitigation as "None."[4] Wiz recommends temporarily disabling HTTP/2.[1] Disabling it removes the precondition both sources describe, but the upgrade is the fix.
Sources
nginx security advisories, nginx.org (accessed November 19, 2026)
Module ngx_http_v2_module, nginx.org documentation (accessed November 19, 2026)
NGINX ngx_http_v2_module vulnerability CVE-2026-123007 (K000163926), F5 (accessed November 19, 2026)
NGINX ngx_http_v2_module vulnerability CVE-2026-123633 (K000163928), F5 (accessed November 19, 2026)
6. Controlling nginx, nginx.org documentation (accessed November 19, 2026)
Originally published at wowhow.cloud
Top comments (0)