· via Cloudflare blog
nginx2root flaws expose millions of nginx servers to crashes and remote code execution
F5 and nginx disclosed five unauthenticated memory-corruption vulnerabilities in nginx, dubbed nginx2root, that can crash workers or enable command execution; Xint chains them with a Linux kernel bug to reach root.

Five flaws, five configurations
On November 18, 2026, F5 and the nginx project disclosed ten vulnerabilities in nginx. According to the Cloudflare blog, five of them, researched by a team called Xint and published under the name nginx2root, are memory-corruption bugs that an unauthenticated attacker can use to crash an nginx worker or, per Xint's research, execute commands on the server. Each bug stands on its own and needs a different setup:
- CVE-2026-123007: integer overflow leading to out-of-bounds heap read/write, requires HTTP/2 enabled, present since 1.9.5 (2015)
- CVE-2026-123633: use-after-free, requires HTTP/2 with proxy_pass to a backend while a worker is low on worker_connections, since 1.19.7
- CVE-2026-121810: use-after-free, requires HTTP/3 (QUIC) with proxy_pass to a backend, since 1.25.0
- CVE-2026-123634: use-after-free, requires the resolver plus proxy_pass with variables plus a subrequest such as mirror or SSI, since 0.6.18
- CVE-2026-113425: use-after-free, affects the mail proxy with proxy on and auth_http, since 0.5.33
F5 rates every one of them High under CVSS 3.1 (8.1) and Critical under CVSS 4.0 (9.2), and notes that code execution requires disabling or bypassing ASLR. For CVE-2026-123007, Xint's write-up shows how to leak memory addresses to do exactly that. Two of the HTTP/2 bugs were also found independently: F5 credits Rapid7, Wiz Research and Ikotas Labs for CVE-2026-123007, and Metnir, Wiz Research and Palo Alto Networks Unit 42 for CVE-2026-123633.
The HTTP/2 bug has the widest reach
CVE-2026-123007 is the headline issue. It needs nothing more than HTTP/2, which is commonly enabled on TLS deployments, and it has shipped in every nginx release since HTTP/2 support arrived in 2015. Xint says it ran the exploit over the public internet against servers on AWS, GCP and Azure. A Shodan query by Xint on November 10, before disclosure, matched 56.7 million hosts identifying as nginx; of the 11.94 million that reported a version, 98.9% fell inside the affected range. As of November 19, Censys's nginx2root exposure tracker counted roughly 8.8 million services running a potentially affected version with HTTP/2 enabled. Cloudflare cautions that version strings alone do not prove a host is vulnerable, because distributions backport fixes.
A kernel bug turns a worker into root
The sixth piece of nginx2root sits in the Linux kernel. Xint found a use-after-free in the SCTP send path, present since Linux 4.15 (2018): while a send operation waits, its SCTP association can be moved to another socket, and the waiting send can then free it. An attacker already running code as the nginx worker can use it to gain root on the host, and Xint demonstrated this from an nginx instance inside a Docker container. Many distributions load the SCTP module automatically the first time a program opens an SCTP socket, so hosts that never use SCTP can still be exposed. As of November 18 there was no kernel fix, and Xint is withholding details until one exists.
Exploitation status and fixes
As of November 19, none of the five CVEs appeared in CISA's Known Exploited Vulnerabilities catalog, and Cloudflare reports no signs of exploitation in the wild. Xint and Wiz Research have not released their exploits. Rapid7 published a proof of concept it describes as suitable only for demonstrating the attack against a freshly started local container, and Cloudflare warns that fake exploits are circulating on GitHub.
The fix is to update to nginx 1.30.7 or 1.31.9, or a vendor build containing the patches, and restart. The same releases close five other bugs, including CVE-2026-118899, an infinite loop while reading a request body over TLS 1.3 that lets an unauthenticated attacker stall a worker in nginx 1.13.0 and later.
Reverse proxies blunt the worst attacks
Cloudflare says its own edge is unaffected: much of its proxy now runs on FL2, written in Rust, and the parts built on its modified nginx fork were reviewed and updated globally before disclosure. For customers, origins reachable only through Cloudflare's HTTP proxy are shielded from the HTTP/2 and HTTP/3 attacks, because Cloudflare terminates every client connection at its edge and opens a separate connection to the origin. The attacker's flow-control, settings, ping and stream-reset frames are consumed at the edge and never forwarded, even when Cloudflare speaks HTTP/2 to the origin, since Cloudflare runs its own flow control on that link. Rapid7 makes the same point in its analysis of CVE-2026-123007. The resolver bug is not fully protected behind the proxy and depends on how proxy_pass is built, and the mail proxy bug is not protected at all. One condition applies throughout: the origin must not be directly reachable, or an attacker can bypass the proxy entirely.
Why it matters
nginx powers a large share of the world's web servers, and the easiest of these bugs to trigger has been shipping for a decade behind a feature most TLS sites enable. Chained with the SCTP kernel flaw, a single unauthenticated request path can escalate from a worker crash to root on the host, including from inside a container. Patching nginx is straightforward, but the kernel gap leaves host-level risk open until a fix lands, and proxy-based protection only holds for origins with no direct path from the internet.
- #nginx
- #security
- #vulnerabilities
- #http2
- #web-servers
- #linux-kernel