November 19, 2026

Exploit Code Is Out for ‘NGINX2Root’: Seven Steps to Find and Fix the Web Server Flaws Hiding in Your Stack

F5 has fixed five critical NGINX flaws, one of them present since 2015, that can let an attacker run code on a web server without logging in. Proof-of-concept code is already public. Here is a step-by-step plan for finding exposed servers, patching them properly and covering the gap until you can.

Close-up of bundles of blue and green network cables running across the front of a dark server rack

NGINX is one of the most widely used pieces of software on the internet. It serves websites, sits in front of business applications as a reverse proxy and ships inside countless appliances and container images. On Wednesday F5, which develops NGINX, released fixes for five critical flaws that researchers have grouped under the name “NGINX2Root,” and public exploit code followed quickly. The guidance below draws on The Hacker News’ detailed write-up and SecurityWeek’s coverage of the release.

Why this one deserves attention today

The headline bug, CVE-2026-123007, is a buffer overflow in NGINX’s HTTP/2 handling that has been in the open-source code since version 1.9.5, released in September 2015. Wiz reported running code on an NGINX server that was only serving ordinary static files, with no backend application and no add-on modules. F5 scores all five flaws 9.2 out of 10 under CVSS 4.0 and says code execution requires ASLR, a standard memory protection, to be off or bypassed. Researchers say that condition is weaker than it sounds: Wiz says two of the HTTP/2 flaws can themselves be used to get around ASLR, and Rapid7’s public exploit does so by leaking memory addresses from the server’s own responses.

As of Wednesday, neither F5 nor the researchers had reported attacks in the wild, and CISA had not added the flaws to its Known Exploited Vulnerabilities list. Censys, however, counts roughly 9 million internet-facing NGINX services running an affected version with HTTP/2 switched on, plus about 103 million more that do not reveal their version.

The walkthrough

  1. List everything that runs NGINX. Think beyond the obvious web server. Reverse proxies in front of line-of-business apps, load balancers, container images, hosting control panels and vendor appliances may all include it. Ask your web host and software vendors directly if you are not sure.
  2. Find out which setups are reachable. Each of the five flaws needs a particular configuration:
    • CVE-2026-123007 (buffer overflow) and CVE-2026-123633 (use-after-free) need HTTP/2 to be enabled.
    • CVE-2026-121810 needs HTTP/3.
    • CVE-2026-113425 affects NGINX acting as a mail proxy.
    • CVE-2026-123634 affects proxies whose upstream server name is built from variables, when a server-side include (SSI) or mirror subrequest looks up that same name.

    To check HTTP/2 on a given server, connect with the openssl s_client command using the -alpn h2 option; if the output reports “ALPN protocol: h2,” HTTP/2 is accepted. If a CDN sits in front of the site, test the NGINX origin server itself, because an origin that only speaks HTTP/1.1 cannot be reached through the HTTP/2 bugs.

  3. Upgrade to a fixed build. Every open-source release from 1.9.5 through 1.31.8 is vulnerable, including the October security releases. NGINX Open Source 1.31.9 (mainline) and 1.30.7 fix all five flaws. On NGINX Plus, 37.3.0.1 and 37.0.8.1 fix four of them; the CVE-2026-123633 fix arrived earlier in 37.2.0.1 and 37.0.7.1. Plus releases R34 through R36 are also affected.
  4. Restart, do not just reload. After upgrading, do a full restart or NGINX’s live binary upgrade. A configuration reload keeps the old executable running, and existing connections can stay on old worker processes.
  5. Watch your Linux distribution’s own packages. The official NGINX Docker images, including Alpine variants, had fixed tags on Wednesday, but Ubuntu and Debian had not yet shipped fixed packages, and their trackers still listed nginx as vulnerable across many releases. Distribution packages can carry a fix without changing the version number NGINX reports, so compare the full package version against the distribution’s advisory rather than trusting what nginx -v reports.
  6. Cover the gap if you cannot patch yet. Turning off HTTP/2 is the temporary measure suggested in guidance from CIRCL, Luxembourg’s computer incident response center. It blocks the two HTTP/2 flaws but not the other three. Turning off HTTP/3 where it is used removes what CVE-2026-121810 needs. F5 lists workarounds for only two flaws: avoid SSI and mirror directives and do not build proxied server names from variables (CVE-2026-123634), and keep mail proxy listeners on trusted networks (CVE-2026-113425). After any change, rerun the openssl test to confirm HTTP/2 is really gone.
  7. Reduce the damage a break-in could do, and look for signs of one. Code run through these bugs inherits the NGINX worker’s permissions; getting full control of the host takes a second privilege-escalation bug. Xint showed one route, chaining CVE-2026-123007 with a Linux kernel bug in SCTP code to escape a Docker container and gain root. On hosts that do not use SCTP, CIRCL’s guidance suggests blocking the module from loading, for example with an “install sctp /bin/false” line in a file under /etc/modprobe.d/ (this does not unload a module that is already loaded). Signs worth checking include NGINX workers launching unexpected programs, workers crashing or leaving core dumps, long-lived HTTP/2 connections with heavy inbound traffic but few logged requests, and unusual outbound connections or DNS lookups. A clean access log is not proof of safety, because HTTP/2 control frames do not normally appear there.

How hard is the attack in practice?

Opinions differ. Rapid7’s proof of concept needs roughly 128 GB of HTTP/2 traffic, and it notes that on links slower than about 300 Mbit/s the transfer would outlast NGINX’s default one-hour connection limit and fail; proxied HTTP/2 connections also tend to break it. The notes on one public proof of concept say it is only suitable for demonstrating the attack against a freshly started Docker container on the same machine. Censys, on the other hand, says it reproduced the underlying behavior in its lab over cleartext HTTP/2 without that heavy traffic and has published its own proof of concept. Rapid7’s method also depends on a memory-copy routine used on x86-64 processors with certain AVX-512 features. Treat “hard to exploit” as a reason to schedule calmly, not a reason to wait.

F5’s same release also patched a set of BIG-IP issues and several lower-severity NGINX flaws; owners of BIG-IP equipment should review those advisories separately.

What this means for your company

Even if you have never installed NGINX yourself, there is a good chance it is running somewhere on your behalf: on your website, in your hosting provider’s stack, inside a cloud application or in a network appliance. This week, ask whoever manages your website and customer-facing systems three questions: Do we run NGINX anywhere? Which version, and does it have HTTP/2 enabled? When will it be on a fixed build, restarted properly? For anything you cannot update quickly, turning off HTTP/2 and closing unused services buys time. Keep web servers running with minimal privileges and up-to-date host operating systems, so one bug does not hand over the whole machine.

If you are not sure where NGINX lives in your environment or who is responsible for patching it, we can help you map it and close the gaps quickly. Reach out to Layer 7 Data Solutions to discuss how we’ve helped companies similar to yours.

About The Author