GBHackers

NGINX2ROOT Exploits 5 Critical NGINX Vulnerabilities to Gain Remote Code Execution and Root Access


Five memory corruption flaws in F5 NGINX Open Source and NGINX Plus, collectively named nginx2root, let remote, unauthenticated attackers run code on exposed web servers.

Disclosed on November 18, 2026, the bugs are CVE-2026-123007, CVE-2026-123633, CVE-2026-123634, CVE-2026-121810, and CVE-2026-113425. F5 rates each 8.1 (High) under CVSS v3.1 and 9.2 (Critical) under CVSS v4.0.

NGINX2ROOT Exploits 5 Critical NGINX Vulnerabilities

Each flaw is an independent attack path that needs a different feature to be enabled. CVE-2026-123007 is a heap buffer overflow in HTTP/2 flow control and needs only HTTP/2. CVE-2026-123633 is a use-after-free in proxied HTTP/2 connections. CVE-2026-123634 is a use-after-free in resolver and subrequest handling. CVE-2026-121810 is a use-after-free in proxied HTTP/3 (QUIC) traffic.

CVE-2026-113425 is a use-after-free in the mail proxy during login and requires no valid credentials. On their own, the flaws give code execution with worker-process privileges where ASLR is disabled or bypassed. This follows a string of recent NGINX remote code execution bugs, including an 18-year-old rewrite module flaw with a public exploit.

The HTTP/2 bug is the biggest concern because its vulnerable code has shipped since HTTP/2 arrived in nginx 1.9.5 in 2015. It affects every official release through 1.31.8. Researchers at Xint found the flaw. They showed that it comes from a flow-control window accounting error. The error pushes the connection window past its 31-bit limit.

An unsigned underflow then turns into an oversized memory copy. Wiz demonstrated remote code execution with an ASLR bypass against official 1.31.7 and 1.31.8 builds, and Rapid7 published a proof of concept for this CVE only. The exploit is not plug-and-play. Rapid7 says it needs roughly 128 GB of HTTP/2 traffic and was built against a single test setup. Censys had seen no exploitation in the wild as of November 18.

Reaching root takes a second flaw. Xint chained the HTTP/2 bug with a Linux kernel vulnerability in the SCTP subsystem, and the chain also escaped a Docker container. No kernel patch exists yet. Whether the chain works depends on kernel modules and container restrictions.

The same subsystem has been abused before in an 18-year-old Linux kernel SCTP flaw that allowed root escalation and container escape. Xint advises blocking SCTP module loading on hosts that do not need it.

Exposure is large. Censys counts 331,242,037 internet-exposed NGINX web properties. About 9.6 million services run an affected version with HTTP/2 enabled. Since tracking began, about 1.08 million exposed services have been fixed, and 590,594 of those were fixed in the last 24 hours. Most servers hide their version, so the real number is higher.

Ubuntu backports fixes without changing the version string, so a version number alone does not show patch status. Censys tracks live figures on its nginx2root exposure dashboard and in its rapid response advisory.

All five flaws are fixed in NGINX Open Source 1.30.7 (stable) and 1.31.9 (mainline). Version 1.31.8 is affected even though its number is higher than 1.30.7. NGINX Plus fixes are 37.0.8.1 and 37.3.0.1. Older branches get no fix.

Turning off HTTP/2 blocks only CVE-2026-123007 and CVE-2026-123633. The other three need separate mitigations. These are fixed upstream groups for resolver paths, removing QUIC listeners for HTTP/3, and restricting access to mail proxies. Native gRPC services that require HTTP/2 will break if you turn it off.

A configuration reload does not replace the running binary, so restart nginx or complete a supported live binary upgrade. Administrators who have dealt with earlier HTTP/2 abuse, such as the HTTP/2 Bomb DoS attack, will recognize the same need to check each listener.

Detection is harder than patching. The HTTP/2 control frames this bug uses are not logged as requests, so access logs will not show an attack. Xint calls processes launched by the nginx worker user are the clearest sign of a successful attack.

Repeated worker exits and unusual connection traffic are weaker signs, because they can have benign causes. Teams that have not yet reviewed their exposure after the earlier Nginx-poolslip memory corruption flaw should audit all NGINX deployments now.

Give your SOC the alert context on the latest threats to stop incidents quickly -> Integrate Threat Intel from 16K SOCs:



Source link