On 18 November 2026, researchers disclosed five flaws in nginx that they call NGINX2Root. F5, which makes nginx, released fixed versions the same day. Researchers have shown the main flaw, CVE-2026-123007, running their own code on servers without signing in.
No attacks had been reported by 19 November. But exploit code is public, and nginx runs in websites, cloud services, containers and many products.
What happened
nginx is a web server that also passes visitors’ requests on to other systems. Xint, a security research firm, reported the flaws to F5 in October.
F5 scores all five at 9.2 out of 10 (CVSS 4.0). Each one needs a particular setting to be switched on:
- CVE-2026-123007: HTTP/2, a common, faster way for browsers to talk to web servers. This is the one most servers will have. It affects every release since 1.9.5, from 2015.
- CVE-2026-123633: HTTP/2, with requests passed on to another server.
- CVE-2026-121810: HTTP/3, a newer version of the protocol.
- CVE-2026-123634 and CVE-2026-113425: less common setups, including nginx used as a mail proxy.
An attacker who succeeds runs commands with the rights of nginx’s worker process. Wiz showed this on a server serving only ordinary files. Xint combined CVE-2026-123007 with a separate Linux kernel flaw to break out of a container and take over the server. That kernel flaw had no fix by 19 November.
Sources differ on how hard this is. F5 says code execution “is only possible if ASLR is disabled or bypassed”, ASLR being a standard memory defence. Wiz says two of the flaws can get around it.
Censys, which maps what is reachable on the internet, counts about 9 million nginx services on the internet on an affected version with HTTP/2 switched on. A further 103 million don’t reveal their version.
The fixed versions are NGINX Open Source 1.30.7 and 1.31.9, and NGINX Plus 37.0.8.1 and 37.3.0.1. F5 lists no fix for NGINX Plus R34 to R36, so those customers need to move to 37.x.
What to do now
The researchers and security firms give the same advice. Find every copy of nginx, including those inside products, appliances, containers and Kubernetes. Update it, starting with what the internet can reach.
Two details matter. Xint warns that reloading nginx’s settings doesn’t replace the running program, so restart it after updating. And Linux distributions may fix their own packages without changing the version number, so check their advisories rather than the number alone.
Where an update must wait, turning off HTTP/2 blocks CVE-2026-123007 and CVE-2026-123633, but not the other three. Tenable lists separate workarounds for those.
The Blueprint requirements this tests:
- RW-DEV-02 A list of the software and services you run (Foundation). “Keep a current list of the software installed on your devices and servers, and of the cloud services you use, showing the version of each where it has one and who looks after it.” Would your list show nginx running inside other products?
- RW-DEV-10 Critical weaknesses fixed within 48 hours (Established). These are critical flaws with a public exploit, so internet-facing nginx falls into this group.
- RW-DEV-20 Third-party code kept free of known flaws (Established). If your own products or container images include nginx, update and rebuild them.
- RW-NET-18 Everything exposed to the internet is known (Established). Knowing which nginx servers the internet can reach tells you what to fix first.
- RW-DET-06 Urgent security warnings acted on (Foundation). Watch for fixes from your Linux distribution and from suppliers whose products carry nginx, and for the kernel fix.
How RiskWitness helps
Where do we run nginx? Assets. Record the servers, containers and appliances that run it, with their owner and where they’re hosted.
Is this already on our register? Risks. Most registers already hold a risk such as “Attackers exploit a known flaw before our fixes are installed”. Review its owner, scores and review date now, and link the affected assets. If you don’t have one, add it.
Which suppliers does it involve? Suppliers. Link the risk to suppliers whose products or hosted services carry nginx, and record when they say their fix is due.
Who is updating what, and by when? Actions. Give each update and restart an owner and a due date. Overdue actions show on the dashboard.
Did it work? Controls. Link your patching control to the risk and map it to RW-DEV-10. Record a test once the updates are in, with its result.
What about servers we can’t update yet? Risk acceptance. Where HTTP/2 is turned off for now, or you’re waiting on a supplier, record the acceptance with a reason and an expiry. If it’s outside tolerance, a named approver signs it off.
Does leadership know? Reports. Share a PDF report or slide deck showing the risk and the actions.
Every step is kept in a history that shows who did what and when, with the old and new values. Nobody can edit or delete it. When leadership, an auditor, a customer or a regulator asks how you handled NGINX2Root, the answer is already written down.
Sources
- NGINX2Root: new remote code execution vulnerabilities in nginx require urgent patching, Xint, 18 November 2026
- F5 patches critical NGINX2Root flaws as PoC exploits go public, SecurityWeek, 19 November 2026
- NGINX2Root FAQ: five vulnerabilities in NGINX (CVE-2026-123007), Tenable, 18 November 2026, updated 19 November 2026
- NGINX2Root: five nginx flaws, led by CVE-2026-123007. How to find, mitigate, and patch now, Qualys, 19 November 2026
- NGINX2Root: five nginx vulnerabilities, starting with CVE-2026-123007, Bitsight, 18 November 2026
