
nginx2root: critical bugs in the server in front of your app
On November 18, 2026, F5 and nginx.org disclosed a batch of security flaws in nginx. Five of them, researched under the name nginx2root, are memory corruption bugs. An attacker who can reach your nginx listener does not need a password. In the worst cases they can run code as the nginx worker process, and researchers showed how that foothold can be pushed further on the host.
nginx sits in front of a huge share of the web: reverse proxies, load balancers, API gateways and static sites. Wiz notes it powers more than 225 million websites. When the front door is broken, a secure app behind it is not enough.
What was disclosed
The two HTTP/2 flaws that matter most for typical public websites are:
| CVE | What it is | Why it matters |
|---|---|---|
| CVE-2026-123007 | Memory corruption in HTTP/2 flow control | Unauthenticated code execution. Affects nginx back to 1.9.5 (2015). HTTP/2 on a listening port is enough. |
| CVE-2026-123633 | Use-after-free when HTTP/2 connections are drained while proxying | Unauthenticated code execution when nginx accepts HTTP/2 and forwards to an upstream. |
Both are rated critical (CVSS 4.0 score 9.2 in vendor reporting). Wiz Research and others independently demonstrated remote code execution, including on ordinary static files served over HTTP/2 with TLS. No custom app code required.
Three more nginx2root issues hit other paths (HTTP/3, a specific resolver and mirror setup, and the mail proxy). Same outcome class: unauthenticated code execution as the worker when the matching features are enabled.
When are you at risk?
You are in the danger zone if all of this is true:
- nginx is reachable from the internet (or from a network an attacker can reach)
- you still run an unpatched version (below 1.30.7 / 1.31.9, or an unpatched nginx Plus build)
- HTTP/2 is enabled on that listener (the usual setup for modern HTTPS sites)
That covers a lot of real deployments: a VPS with nginx in front of Node, PHP, WordPress or Docker; a reverse proxy that terminates TLS; even a static site served only by nginx. Wiz showed RCE on servers that only hosted ordinary static files over HTTP/2 with TLS. Your application code can be perfect and you can still be vulnerable, because the bug is in nginx itself.
You are less exposed if HTTP/2 is off, if a patched edge (for example a CDN that terminates TLS for you) sits in front and your origin is not directly reachable, or if you already upgraded. You are not safe just because "we only have a brochure site" or "we have a firewall". The attack is a crafted network request to nginx. No login, no admin panel, no weak password required.
What can a hacker do with this?
Once they run code as the nginx worker, they are on your machine. From there they can typically:
- Take over the website: change pages, inject malware or phishing, redirect customers, or shut the site down
- Steal secrets nginx can see: TLS material, environment variables, config files, API keys, database credentials used by apps on the same host
- Read or modify app data: reach databases, uploads and admin panels sitting behind the proxy
- Spy on traffic: because nginx terminates HTTPS, a compromised worker can see requests and responses that look encrypted to the outside world
- Stay persistent: install backdoors, crypto miners or bots that survive a simple "restart nginx"
- Escalate further: the nginx2root research also showed chaining into a Linux kernel bug to get root on the host, including from inside Docker in some setups. Worker access is already bad. Root means the whole server is theirs.
In plain terms: this is not a minor bug that only crashes a page. It is a path from "anyone on the internet" to "control of the box that serves your business".
Who is affected (versions)
- nginx Open Source mainline through 1.31.8, and stable 1.30.x through 1.30.6 (exact ranges depend on the CVE).
- nginx Plus releases in the R34 to R37 range listed in the vendor advisory.
Fixed releases include 1.30.7 (stable) and 1.31.9 (mainline). nginx Plus users should move to the patched builds F5 published (for example 37.0.8.1 / 37.3.0.1).
If you cannot patch immediately, turn off HTTP/2 on exposed listen directives. That mitigates the two main HTTP/2 RCEs until you can upgrade.
Why this hits businesses hard
Most teams treat nginx as plumbing. It is "just the proxy". Attackers treat it as the highest-value target on the box: it terminates TLS, sees every request, and often runs with enough privilege to reach your apps, configs and secrets.
nginx2root is also a reminder that shared infrastructure debt is business risk. A server left on an old nginx package for "stability" can sit vulnerable for years. CVE-2026-123007 reaches back more than a decade of releases.
CERT-EU advised patching as soon as possible, especially for internet-facing appliances. At disclosure time, exploitation in the wild had not been broadly reported, but working research demos exist. That gap does not last.
What you should do now
- Check your nginx version (
nginx -v) on every public edge and reverse proxy. - Upgrade to 1.30.7+ or 1.31.9+ (or the matching nginx Plus fix).
- Until then, disable HTTP/2 on public listeners.
- Prefer managed edges and modern, patched stacks over "set and forget" VPS images that never get security updates.
Own the stack you put online
At codebind we build and host applications on modern, maintainable stacks, with infrastructure you can patch and control. If your site still sits behind an unmaintained nginx box, or an aging PHP and WordPress stack that compounds the same risk, use Refactor Website on our homepage and we will quote a migration to something you actually own.
Sources: nginx security advisories, Wiz Research, Cloudflare, CERT-EU, BleepingComputer.