Consulting & Audit Cybersecurity & IT security

NGINX2Root: five NGINX vulnerabilities, including CVE-2026-123007

EG Elisa GUYON · · 16 min read

On 18 November 2026, F5 published patches for five vulnerabilities affecting NGINX Open Source and NGINX Plus. Grouped under the name “NGINX2Root”, they cover several server functions, including HTTP/2, HTTP/3, DNS resolution and the mail proxy.

The most widely discussed flaw, CVE-2026-123007, exploits an error in HTTP/2 flow control. It can cause memory corruption, crash an NGINX process and, under certain conditions, allow unauthenticated remote command execution. A public PoC is available for this vulnerability.

The five flaws do not affect every configuration in the same way. Exposure depends on installed versions and the features enabled on the server. Here is what to check and which measures to take.

Before reviewing each CVE, it helps to understand why these vulnerabilities are grouped under one name and what that means for IT teams.

NGINX2Root: what this security alert covers

NGINX2Root designates a family of five vulnerabilities found in NGINX Open Source and NGINX Plus. They affect several server components: HTTP/2 and HTTP/3 connection handling, domain name resolution and the mail proxy. F5 published its advisories and patches on 18 November 2026.

NGINX is used to serve websites, distribute traffic or forward requests to applications and backend servers. A vulnerability in this software can therefore affect very different setups. The mere presence of NGINX is not enough to decide whether a server is exposed. You must check its version and enabled features.

That distinction matters when assessing risk. The alert covers a coherent family of recent flaws, distinct from other NGINX vulnerabilities already published. To place this incident in a broader approach, see Complianz System’s cybersecurity hub and the dedicated cybersecurity audit page.

Among the five vulnerabilities, CVE-2026-123007 draws particular attention because it affects HTTP/2 handling and may impact servers that accept this type of connection.

CVE-2026-123007: memory corruption in HTTP/2

The flaw comes from an error in HTTP/2 control-window calculation. That mechanism regulates how much data the server may send to a client. In some circumstances, the window can exceed its allowed limit. NGINX may then handle an incorrect amount of memory and cause heap corruption.

A remote attacker does not need to authenticate to attempt exploitation. The main condition is that the server accepts HTTP/2 connections on an accessible address. Exploitation can crash a worker process. Under certain technical conditions, it may also allow reading or modifying memory, then remote code execution with that process’s privileges.

Severity varies by scoring scale. CVE-2026-123007 scores 8.1 under CVSS 3.1 (High) and 9.2 under CVSS 4.0 (Critical). These scores describe potential severity; they do not mean every NGINX server is equally exploitable.

The fix is available in recent NGINX versions. Until they are installed, disabling HTTP/2 on exposed listeners can reduce risk for this CVE. That measure remains temporary and does not address the other vulnerabilities in the family.

The other four vulnerabilities are not all based on HTTP/2. Their exploitation depends on particular modules and settings, which is why the real server configuration must be checked.

The five NGINX vulnerabilities and their exploitation conditions

CVE Component affected Configuration to check
CVE-2026-123007 HTTP/2 HTTP/2 enabled on an NGINX listen point
CVE-2026-123633 HTTP/2 and upstream connections HTTP/2 enabled and requests forwarded to an upstream server
CVE-2026-121810 HTTP/3 and QUIC HTTP/3 enabled, with connections to an upstream server
CVE-2026-123634 DNS resolution and proxy Dynamic resolution with resolver, proxy_pass destination built from the request, and a matching mirror or SSI subrequest
CVE-2026-113425 Mail proxy NGINX configured as a mail proxy with auth_http

CVE-2026-123633 is a use-after-free vulnerability. It appears in a scenario where NGINX closes an active HTTP/2 connection while communicating with an upstream server. CVE-2026-121810 follows a similar mechanism but concerns HTTP/3 and QUIC. Because HTTP/3 support is tied to particular versions and configurations, its activation must be checked separately.

CVE-2026-123634 affects DNS resolution handling in some proxy configurations. Risk depends on the combination of directives, not simply on the presence of an NGINX proxy. CVE-2026-113425 targets installations that use NGINX as a mail proxy and rely on external authentication checks.

These differences change how exposure is assessed. Version detection alone does not replace reviewing active modules and directives. A penetration test can complement that review when you need to verify, within an authorised scope, whether a weakness is truly exploitable.

Exploitation conditions help identify configurations to review. To confirm whether a server remains vulnerable, you must also compare its version with the matching patch.

Which NGINX versions are affected?

For NGINX Open Source, patches in this series are included from versions 1.31.9 and 1.30.7. Earlier branches may be affected, but the exact vulnerable range varies by CVE and features in use. A version number alone is therefore not always enough to conclude that a server is exposed.

For NGINX Plus, the fixed versions indicated by F5 are 37.3.0.1 for the recent 37.x branch and 37.0.8.1 for the 37.0 branch. Deployments on another branch should consult the F5 advisory for their product and version.

Another difficulty concerns Linux distributions and hosting panels. Debian, Ubuntu, Red Hat or a host may ship a security fix in their own package without matching the version number published by the NGINX project. Check the security advisory from the OS or service vendor rather than relying only on the version shown by nginx -v.

Teams should inventory NGINX instances, record version and origin, then check available fixes with the relevant vendor. They should also identify Internet-facing services and enabled directives, especially HTTP/2, HTTP/3, dynamic proxying and the mail proxy. An IT security policy helps formalise who checks these components, how often, and how urgent updates are tracked.

Beyond affected versions, publication of a proof-of-concept changes the attention due to CVE-2026-123007. A technical demonstration must still be distinguished from attacks observed on production servers.

Public PoC: what is the concrete risk?

On 18 November, Rapid7 published a demonstration of CVE-2026-123007 exploitation. It shows unauthenticated remote code execution on a test environment using NGINX 1.31.8, HTTP/2 and TLS. That proof of concept concerns this specific CVE. It does not show that the other four family vulnerabilities are exploited the same way.

Those constraints do not eliminate the risk. Public code makes analysis easier and may help other researchers or attackers adapt their own methods. Organisations should therefore treat the patch as a priority, especially if HTTP/2 is enabled on an exposed server.

At the time advisories were published, CERT-EU had not reported exploitation of these five flaws in the wild. That absence of reporting does not prove no attempts occurred. It only describes the information available on that date. Prevention should combine updates, configuration checks and the ability to spot abnormal behaviour, as recommended in a cyber-attack protection approach.

Once affected versions and features are identified, the priority is to install patches suited to the distribution or NGINX edition in use.

How to fix the NGINX flaws

For NGINX Open Source, the fixed versions indicated are 1.31.9 and 1.30.7. For NGINX Plus, F5 indicates versions 37.3.0.1 and 37.0.8.1 depending on the branch. Teams using a package from their Linux system, host or admin panel should apply the fix published by that vendor. The version number may differ if the fix was backported into the package without an upstream upgrade.

After installation, check the configuration with nginx -t, then restart or upgrade the service per the vendor procedure. A simple configuration reload does not guarantee the corrected binary is the one running. Then verify the active version and service status.

If the update cannot be applied immediately, disabling HTTP/2 on exposed listeners can temporarily reduce risk for CVE-2026-123007 and CVE-2026-123633. That measure does not fix CVE-2026-123634, CVE-2026-121810 or CVE-2026-113425. For those flaws, exposure depends on other features such as HTTP/3, certain proxy configurations or the mail proxy.

Disabling features can affect the service delivered to users. Test it, document it and treat it as provisional. The priority remains installing a confirmed patch for every product and branch deployed.

Deploying the patch does not fully close the response. You must also confirm every instance is patched and look for signs of abnormal activity.

What to check after the update

Start by revisiting the NGINX server inventory and comparing it with the list of patched systems. Check production instances, test environments, virtual machines and services embedded in a hosting panel or application stack. A patch applied on one server does not protect forgotten instances.

Confirm the NGINX process is using the corrected binary and that the service restarted correctly. Then review HTTP/2, HTTP/3, DNS proxy and mail configurations. If a feature was temporarily disabled, record the change and plan to reassess it after the patch is installed.

Review error and monitoring logs for unusual worker process crashes. A message that a worker stopped on a signal may match an exploitation attempt, but alone does not prove successful code execution. Correlate it with other signals, including process-creation logs, unusual outbound connections and observed changes on the server.

If these checks reveal indicators of compromise, treat the event as a security incident and retain logs useful for analysis. Finally document patched versions, temporary measures and any investigations in your security tracking. That helps verify remediation covers the full scope.

The NGINX2Root alert brings together five vulnerabilities that affect different components and configurations. In conclusion, the response must therefore combine updates, technical verification and incident follow-up.

Key takeaways on NGINX2Root vulnerabilities

CVE-2026-123007 is the most widely discussed flaw in this series. It concerns HTTP/2 and has a public proof of concept. The other four vulnerabilities notably affect HTTP/2 with proxying, HTTP/3, DNS resolution and the mail proxy. Their exploitation depends on features actually enabled on each server.

Patches published by F5 and the NGINX project should be applied quickly to the affected versions and packages. Temporarily disabling HTTP/2 can reduce exposure to two of the five vulnerabilities, but it does not replace a full update. Teams must also account for packages adapted by Linux distributions and hosting providers.

Finally, the name NGINX2Root does not mean these flaws automatically grant root access. Executed code may have the privileges of the NGINX process, depending on server configuration. For organisations, the challenge is knowing exactly where NGINX is installed, which features are used, whether the patch is active and whether logs show unusual activity.

FAQ on the NGINX2Root vulnerabilities

Short answers on NGINX2Root, CVE-2026-123007, the public PoC, fixed versions and what “root” really means here.

What is NGINX2Root?

NGINX2Root is the name given to a family of five vulnerabilities affecting NGINX Open Source and NGINX Plus. They cover several functions, including HTTP/2, HTTP/3, DNS resolution and the mail proxy. Affected configurations vary by CVE.

Does CVE-2026-123007 affect every NGINX server?

No. This flaw concerns servers that accept HTTP/2 connections and run a vulnerable version. The other vulnerabilities depend on distinct configurations, such as HTTP/3, certain proxy directives or the mail proxy.

What does CVE-2026-123007 allow?

It causes an error in HTTP/2 flow control that can lead to heap memory corruption. An unauthenticated remote attacker can crash a worker process and, under certain conditions, execute code with that process’s privileges.

Is there a public PoC for NGINX2Root?

Yes, a public PoC was published for CVE-2026-123007. It demonstrates remote code execution in a specific test environment. That does not mean all five vulnerabilities have an equivalent public PoC.

Are NGINX2Root vulnerabilities exploited in the wild?

At the time advisories were published, CERT-EU had not reported exploitation in the wild. That reflects reports available on that date. It does not guarantee that no attempts occurred.

Does disabling HTTP/2 protect against all five vulnerabilities?

No. Disabling HTTP/2 can temporarily reduce exposure to CVE-2026-123007 and CVE-2026-123633. It does not fix flaws related to HTTP/3, DNS resolution or the mail proxy. Updating remains the priority measure. See also cyber-attack protection.

Which NGINX versions fix these vulnerabilities?

For NGINX Open Source, the fixed versions indicated are 1.31.9 and 1.30.7. For NGINX Plus, patches are available in versions 37.3.0.1 and 37.0.8.1, depending on the branch. Linux distributions and hosts may provide their own patched packages.

Does NGINX2Root directly grant root access to the server?

Not necessarily. Code execution may run with the privileges of the NGINX worker process. The access level obtained depends on rights granted to that process and on server configuration.