Security Intelligence
November 19, 2026 | Security intelligence
NGINX2Root: Path to code execution and root compromise
At a glance:
-
F5 fixed 10 vulnerabilities in NGINX Open Source and NGINX Plus, including five critical flaws disclosed as part of the NGINX2Root research that may lead to code execution under specific conditions.
-
Researchers demonstrated code execution against affected NGINX deployments, and public proof-of-concept code for CVE-2026-123007 was released on November 18, 2026.
-
Internet-facing NGINX deployments with HTTP/2 enabled are the most relevant systems to review. This includes reverse proxies, Kubernetes ingress controllers, load balancers, and application gateways that sit in front of websites, applications, and APIs.
Threat summary
On November 18, 2026, F5 released an out-of-band security update fixing 10 vulnerabilities in NGINX Open Source and NGINX Plus.
Five of those form “NGINX2Root,” the name Xint gave to a set of flaws that demonstrated paths from internet-accessible NGINX services to code execution. In some cases they also led to root access on the underlying host, when combined with a separate, unpatched Linux kernel vulnerability.
Researchers from Rapid7, Wiz Research, Ikotas Labs, and Metnir independently reported overlapping findings.
The other five vulnerabilities are configuration-dependent issues affecting specific NGINX features. Their reported impact includes denial-of-service, client certificate bypass, access to internal resources, and, in one uncommon configuration, a buffer overflow with theoretical code execution. They are rated medium to high severity, with CVSS v4 scores from 6.3 to 8.7.
NGINX2Root
NGINX is one of the most widely deployed web servers and reverse proxies in enterprise environments. It is also commonly embedded in Kubernetes ingress controllers, container images, application gateways, commercial software, and security appliances, which can make affected deployments hard to identify.
The NGINX2Root research covers five vulnerabilities across four components:
-
HTTP/2: CVE-2026-123007 and CVE-2026-123633
-
HTTP/3: CVE-2026-121810
-
DNS resolver: CVE-2026-123634
-
Mail proxy: CVE-2026-113425
The HTTP/2 vulnerabilities received the most attention because they are subject to a public exploit code. These vulnerabilities affect NGINX deployments that expose HTTP/2 services to untrusted networks.
CVE-2026-123007 applies to systems with HTTP/2 enabled, while CVE-2026-123633 only affects deployments that use NGINX as a reverse proxy under specific connection-management conditions.
F5 describes both vulnerabilities as memory corruption flaws that can cause worker process crashes and, under certain circumstances, may lead to remote code execution.
Rapid7 published a proof-of-concept (PoC) code repository containing a worker-process crash trigger and a full exploit that executed commands against NGINX 1.31.8 running in a Docker container. The repository notes state the exploit is suitable only against a freshly started Docker container on the same machine.
Other repositories claiming to exploit CVE-2026-123007 have appeared on GitHub, and reports of fake exploits have been circulating.
The full exploit depends on conditions beyond the presence of a vulnerable server:
- Approximately 128 GB of HTTP/2 traffic over a single connection, delivered within NGINX's default one-hour connection lifetime
- Direct HTTP/2 access to the target server
- An exploit calibrated for the target's exact NGINX build and C library
- A target with little other traffic, since concurrent requests can cause the attempt to fail
- An x86-64 processor with AVX2 or AVX-512; the attack fails against the older SSE memory-copy routine, but nearly all current server processors support AVX2
The exploit defeats Address Space Layout Randomization (ASLR) on its own by leaking memory addresses through the same flaw, so ASLR-enabled targets are not protected.
Wiz Research reported code execution, with ASLR defeated, against official NGINX 1.31.7 and 1.31.8 builds and withheld its exploit code. Xint demonstrated an attack chain that combined CVE-2026-123007 with an unpatched Linux kernel vulnerability in Stream Control Transmission Protocol (SCTP) handling to obtain root access on the underlying host, including from inside a Docker container. Xint released demonstration videos and withheld the final steps needed to reproduce its exploits.
F5 assigned all five a CVSS v4 score of 9.2. The advisory describes the expected outcome as memory corruption that may cause worker process crashes and potentially remote code execution. No exploitation in the wild or threat actor activity targeting these vulnerabilities had been reported as of November 18, 2026.
Analysis
NGINX is frequently deployed at the front of web applications and APIs, making it one of the first systems exposed to external traffic. Internet-facing deployments running affected versions with HTTP/2 enabled are the most likely to be affected.
Researchers demonstrated code execution against a server delivering only static content, so exploitation does not require a vulnerable application behind NGINX. Reverse proxy deployments also need review because CVE-2026-123633 only applies to those configurations.
Identifying affected deployments may be more difficult than expected. NGINX is frequently embedded in container images, security appliances, commercial software, and Kubernetes ingress controllers rather than deployed as a standalone web server.
HTTP/2 is not enabled by default in NGINX, but some widely deployed platforms enable it automatically. One example is the community ingress-nginx controller, which has reached end of life and will not receive fixes for these vulnerabilities.
Researchers demonstrated that code execution is possible, but the publicly available exploit was developed for a specific test environment rather than a typical production deployment. The PoC relies on sustained bandwidth to a single connection, direct HTTP/2 access to the server, and compatibility with the target NGINX build. Rapid7's analysis also found that memory corruption can occur before the exploit reaches the demonstrated code-execution path. In practice, this means service disruption may be easier to achieve than full compromise using the published exploit.
Traditional web application firewall (WAF) rules are unlikely to stop the NGINX2Root HTTP/2 vulnerabilities because the flaws occur during HTTP/2 processing rather than in the application layer. The demonstrated attacks can also resemble normal long-lived connections.
Deployments that terminate Transport Layer Security (TLS) and HTTP/2 on a WAF, content delivery network (CDN), load balancer, or reverse proxy may be less exposed because attacker-controlled HTTP/2 frames never reach NGINX. Cloudflare stated that customers using its proxied HTTP service are protected because HTTP/2 and HTTP/3 connections terminate at Cloudflare's edge before traffic is forwarded to the origin. Similar protection may be available from other platforms that terminate HTTP/2 rather than passing it through unchanged.
The effectiveness of this protection depends on the architecture. If threat actors can reach the origin server directly, or if the front-end device uses TLS passthrough rather than terminating HTTP/2, the origin remains exposed. These controls can reduce risk while remediation is underway, but they do not replace upgrading NGINX or disabling HTTP/2 on affected systems.
In the most complete attack chain demonstrated during the research, code execution in the NGINX worker process was followed by root access on the underlying host through a separate Linux kernel vulnerability. The impact depends on the role of the affected server. NGINX deployments acting as reverse proxies, Kubernetes ingress controllers, load balancers, and application gateways often handle authentication traffic, APIs, internal applications, and connections to backend systems. Compromise of those systems could expose sensitive traffic, credentials, and access to downstream services.
Mitigations
To remove exposure to all 10 vulnerabilities, upgrade:
-
NGINX Open Source to 1.30.7 or 1.31.9
-
NGINX Plus to 37.0.8.1 or 37.3.0.1
-
NGINX Ingress Controller to 5.8.0
-
NGINX Gateway Fabric to 2.9.0
Restart NGINX after upgrading, because a configuration reload leaves the old executable running. Migrate NGINX Plus R33 through R36 to a supported 37.x release, as those branches do not receive fixes. For Linux distribution packages, compare package release numbers against vendor advisories, because backported fixes often retain older version numbers.
Identify NGINX deployments across servers, container images, Kubernetes clusters, and appliances. Review configurations using nginx -T and determine whether HTTP/2, HTTP/3, resolver functionality, reverse-proxy directives, or the mail proxy module are enabled. This helps identify which NGINX2Root vulnerabilities apply to each deployment and which systems require priority remediation.
Monitor vendor advisories for updates that incorporate the fixed NGINX releases, as some products maintain their own NGINX packages and may require vendor-specific updates rather than a direct NGINX upgrade.
Where upgrades are delayed, disable HTTP/2 functionality to remove exposure to CVE-2026-123007 and CVE-2026-123633. Organizations running the community ingress-nginx controller can disable HTTP/2 through the controller configuration but may also need to plan migration to a supported alternative because no fixes will be released. Remove unnecessary QUIC listeners, restrict mail proxy access to trusted networks, and review resolver configurations handling user-influenced hostnames. On Linux systems that do not use SCTP, block the sctp kernel module to remove the privilege-escalation path demonstrated by researchers.
Monitor NGINX hosts for worker processes spawning shells or unexpected child processes, which are among the clearest indicators of successful exploitation. Review NGINX error logs for repeated worker crashes, and monitor network telemetry for unusually large, long-lived HTTP/2 connections. Because these attacks generate little useful evidence in standard web access logs, process telemetry and network visibility are the more valuable detection sources.
Avoid running unverified exploit code on production systems, as malicious or modified tools can introduce risk beyond the vulnerability being tested. Prefer vendor guidance, controlled validation methods, and trusted security testing tools when assessing exposure.
Stay on top of emerging threats like this.
Sign up to receive a weekly roundup of our security intelligence feed. You'll be the first to know of emerging attack vectors, threats, and vulnerabilities.

