A severe, decade-old security vulnerability residing within the HTTP/2 implementation of NGINX has been discovered, enabling unauthenticated attackers to achieve remote code execution on any server running vulnerable versions of the widely used web server software with HTTP/2 enabled.

Tracked as CVE-2026-123007 and dramatically dubbed "nginx2root" by security researchers, the critical flaw has silently lurked within NGINX’s open-source codebase for over ten years. According to official documentation and technical disclosures, the vulnerability was originally introduced on September 22, 2015, with the release of version 1.9.5. F5, the enterprise technology company that oversees NGINX, moved swiftly to remediate the crisis by releasing patched versions 1.31.9 and 1.30.7.

The discovery highlights systemic risks embedded deeply within heavily relied-upon internet infrastructure. Security researchers from Wiz demonstrated that they could successfully bypass Address Space Layout Randomization (ASLR)—a common operating system protection mechanism designed to prevent memory corruption exploitation—to execute arbitrary code through the flaw during testing phases. Similarly, researchers at Xint confirmed that successful exploitation required nothing more than a publicly reachable and active HTTP/2 port, eliminating the need for valid user credentials or prior system access.

When code is successfully executed via this vulnerability, it initially inherits the limited privileges and permissions of an NGINX worker process. However, cybersecurity analysts at CIRCL, the Computer Incident Response Center Luxembourg, pointed out that escalation to full root privileges on the underlying host machine is achievable if attackers manage to leverage a secondary vector. In empirical tests conducted by Xint, code running initially within the constrained sandbox of an NGINX worker process managed to achieve complete root access on the host by chaining the attack with a separate, distinct vulnerability residing within the Linux kernel’s SCTP (Stream Control Transmission Protocol) code. Notably, this escalation path remained viable even when the NGINX instance was safely isolated inside a standard Docker container.

Despite the profound severity of the vulnerability, industry watchdogs, vendors, and participating research teams, including Wiz and Rapid7, have confirmed that there are currently no documented instances of CVE-2026-123007 being actively exploited in the wild. As of the initial disclosure, the Cybersecurity and Infrastructure Security Agency (CISA) had not yet added the bug to its heavily monitored Known Exploited Vulnerabilities catalog.

To assist security professionals and system administrators in verifying and testing their environments, a proof-of-concept exploit script has been made publicly available on GitHub. However, the accompanying documentation explicitly emphasizes that the proof-of-concept code is constructed exclusively for laboratory demonstration against a newly initialized Docker container residing on the exact same local machine, rather than for broader network operations.

10-Year-Old NGINX HTTP/2 Flaw Lets Unauthenticated Attackers Execute Code on Servers

The scope of affected software is vast. Every single open-source release spanning from version 1.9.5 through version 1.31.8 is inherently vulnerable. This extensive list includes the recent 1.31.8 release from October 21, as well as prior security releases issued on October 14, namely versions 1.31.7 and 1.30.6. Enterprise deployments running NGINX Plus are similarly impacted, specifically touching 37.x releases prior to the newly patched builds, alongside releases ranging from R34 to R36.

Alongside CVE-2026-123007, the vendor’s patch rollout addresses four additional severe vulnerabilities that F5 warns can likewise culminate in remote code execution under specific ASLR conditions. Each of these five vulnerabilities has been assigned a maximum severity score of 9.2 out of 10 under the CVSS 4.0 framework.

Each of these vulnerabilities demands distinct pre-existing conditions and specific server configurations to be reachable by an attacker. For instance, CVE-2026-123007 requires HTTP/2 to be explicitly turned on. A separate issue, tracked as CVE-2026-123633, requires HTTP/2 to be active while incoming requests are passed along to a backend server precisely at a moment when the NGINX worker is running low on available worker connections. Another bug, CVE-2026-123634, requires a proxied server name built dynamically from variables, combined with either a server-side include or a mirror subrequest targeting a location that resolves the exact same name. Meanwhile, CVE-2026-121810 is tied directly to the enablement of HTTP/3, and CVE-2026-123635 impacts instances where NGINX is deployed explicitly as a mail proxy.

Administrators wishing to verify whether their server configuration exposes them to the HTTP/2 vectors can check their configuration files for listen directives that include the explicit http2 parameter, or instances where http2 on has been globally defined. External verification can also be conducted by running OpenSSL client commands against the target server and inspecting the Application-Layer Protocol Negotiation output. When evaluating infrastructure sitting behind a Content Delivery Network, security teams are advised to test the NGINX origin server directly, as origin infrastructure configured exclusively to accept HTTP/1.1 traffic cannot be targeted via HTTP/2 attack vectors.

Fixes, Workarounds, and Signs of Attack

The newly released NGINX versions 1.31.9 and 1.30.7 provide definitive fixes for all five critical flaws outlined in the advisories. For commercial deployments using NGINX Plus, builds 37.3.0.1 and 37.0.8.1 incorporate fixes for four of the flaws, while the resolution for CVE-2026-123633 was integrated earlier into builds 37.2.0.1 and 37.0.7.1.

System administrators are strongly cautioned that completing an upgrade requires either a full restart of the NGINX service or the execution of its native live binary upgrade procedure. Performing a standard configuration reload alone is insufficient, as it leaves the legacy executable running and allows existing connections to persist indefinitely on old worker processes. Official Docker container images, including variants built on Alpine Linux, have already been updated with tags pointing to versions 1.31.9 and 1.30.7.

10-Year-Old NGINX HTTP/2 Flaw Lets Unauthenticated Attackers Execute Code on Servers

Enterprise distributions such as Ubuntu and Debian require careful package tracking. Because Linux distribution maintainers frequently backport security patches into older package versions without incrementing the primary version string reported by standard command-line flags, administrators must inspect the full package version number against official security trackers rather than relying solely on default version queries.

Official vendor guidance offers direct configuration workarounds for only a subset of the vulnerabilities. For the resolver-related flaw, administrators can mitigate risk by avoiding server-side include and mirror directives and by refraining from constructing proxied server names out of variables. For the mail proxy flaw, restricting listeners strictly to trusted internal networks is advised.

As a temporary defensive posture prior to completing full binary upgrades, incident response agencies such as CIRCL recommend temporarily disabling HTTP/2 entirely across public-facing servers. This blocks the primary pathways for CVE-2026-123007 and CVE-2026-123633, though it does not neutralize the remaining vulnerabilities. Similarly, disabling HTTP/3 where applicable removes the necessary conditions for CVE-2026-121810. For installations utilizing vulnerable versions prior to modern baseline updates, administrators can modify their server blocks to turn off HTTP/2 and strip the protocol flag from listening directives, followed by a thorough service restart and verification check.

To safeguard against the secondary kernel escalation vector involving SCTP while waiting for operating system-level patches, security advisories recommend blocking the SCTP kernel module from loading entirely on systems that do not rely on the transport protocol. This can be achieved by appending an explicit block rule to configuration files within the system modprobe directory, though administrators note this action does not automatically unload modules that are already active in memory.

When hunting for potential indicators of compromise, incident responders look for unusual anomalous behavior, such as NGINX worker processes unexpectedly spawning unauthorized child programs, workers terminating abnormally or leaving behind core memory dumps, long-running HTTP/2 connections characterized by exceptionally heavy incoming data transfers alongside unusually low logged request counts, and unexpected outbound network connections or DNS queries. Security experts emphasize that clean access logs do not definitively rule out an attack, given that low-level HTTP/2 control frames often bypass traditional request logging mechanisms.

Technical write-ups published by Rapid7 indicate that executing the specific memory-corruption attack requires a substantial volume of data traffic, estimating an intake of roughly 128 gigabytes of HTTP/2 traffic delivered over a continuous multi-minute window. On network links running slower than approximately 300 megabits per second, data transfers would inevitably breach NGINX’s default one-hour connection timeout threshold, causing the exploit attempt to fail. Furthermore, the exploitation method relies heavily on specific glibc memory-copy routines natively selected on x86-64 server processors equipped with advanced instruction set features such as avx512f and avx512vl.

Independent validations conducted by Wiz and Xint further confirmed that the exploitation vectors function reliably across major cloud computing platforms, including Amazon Web Services, Google Cloud Platform, and Microsoft Azure, reinforcing the urgent need for global infrastructure updates.