Two critical NGINX bugs let attackers run code remotely

Published

F5 has shipped fixes for two critical flaws in NGINX that can let a remote attacker with no credentials run code on the server, putting every website and API sitting behind it at risk. Both bugs live in how NGINX handles HTTP/2 traffic, and one of them has been in the code since version 1.9.5, released in 2015.

The vulnerabilities, CVE-2026-123633 and CVE-2026-123007, each carry a CVSS 4.0 score of 9.2. They were found independently by five teams: Wiz Research, Ikotas Labs, Metnir, Rapid7 and Xint. Wiz's Nir Ohfeld and Tomer Schwartz describe both flaws in Wiz's write-up, while Rapid7 senior principal researcher Stephen Fewer published a detailed technical analysis of CVE-2026-123007 alongside F5's coordinated release on 18 November.

The bug is in NGINX, not your app

NGINX is the front door for a huge share of the web, handling HTTPS and routing requests to the applications behind it. Wiz stresses that because the flaws sit in NGINX itself, an application with perfectly secure code can still be exposed. A single compromised NGINX box often fronts several services at once.

Both bugs require HTTP/2 to be enabled. Beyond that, their conditions differ:

Wiz says it confirmed remote code execution with an ASLR bypass on official NGINX 1.31.7 and 1.31.8 builds. Rapid7 built its own proof of concept against a Docker image of 1.31.8. In both cases, code runs with the permissions of the NGINX worker process; Rapid7 notes that execution as the master process or as root has not been established.

Hard to pull off, for now

Rapid7's analysis puts real limits on CVE-2026-123007. Its exploit needs roughly 128 GB of HTTP/2 traffic: about two minutes on a 10 Gbit/s link, about seventeen on 1 Gbit/s, and on anything slower than around 300 Mbit/s it runs into NGINX's default one-hour connection lifetime and fails. A proxy in front of NGINX will likely not relay the precise frames the attack needs, and Rapid7's technique depends on a specific x86-64 glibc memory-copy routine selected on CPUs with AVX-512 support. Neither firm is releasing its exploit.

Wiz cautions against reading too much comfort into that. It points to PHP2Shell, which it reported last month, where attacks started against packaged products that ship the exact same build and configuration to many customers. Wiz also says it found both NGINX bugs using its AI code-scanning harness, and argues that AI is eroding the complexity that once kept flaws like these hidden. This follows a separate NGINX release in October that fixed five flaws.

Upgrade to 1.30.7 or 1.31.9, or turn HTTP/2 off

There is no report of exploitation in the wild. But with five teams landing on the same bugs independently, it would be unwise to assume nobody else has. The window between disclosure and the first working attack keeps shrinking, and NGINX sits on enough of the internet to make it worth somebody's effort.

This briefing is provided by IntelFusions for informational and defensive purposes only. It is based on sources assessed to be reliable at the time of writing, and analytic judgments carry the confidence levels indicated. Indicators of compromise are defanged; re-arm them only in controlled environments. IntelFusions is not affiliated with the organizations named and makes no warranty as to completeness or accuracy.

Read the full analysis on IntelFusions