CVE-2026-123633 and CVE-2026-123007: Critical NGINX HTTP/2 Vulnerabilities

CVE-2026-123633 and CVE-2026-123007: Critical NGINX HTTP/2 Vulnerabilities


NGINX is one of the most widely deployed components in modern web infrastructure. It serves as a web server, reverse proxy and load balancer, handling incoming traffic before forwarding requests to backend applications and APIs.

A vulnerability in this layer can have consequences beyond a single application. When multiple websites and services depend on the same NGINX deployment, compromising that entry point can potentially expose several downstream services.

Two critical vulnerabilities, CVE-2026-123633 and CVE-2026-123007, have now been disclosed in NGINX Open Source and NGINX Plus. Both carry a CVSS 4.0 score of 9.2 and involve memory-safety weaknesses in HTTP/2 processing.

The vulnerabilities were independently discovered by researchers from Wiz Research, Ikotas Labs, Metnir, Rapid7 and Xint, and reported to F5 through coordinated disclosure.

CVE-2026-123633 – HTTP/2 Use-After-Free

The first vulnerability involves a use-after-free condition in NGINX’s HTTP/2 connection handling.

A use-after-free occurs when an application continues accessing memory after that memory has already been released. Depending on how the freed memory is subsequently reused, this can result in memory corruption, application crashes or potentially arbitrary code execution.

In this particular case, the vulnerability occurs under specific reverse-proxy configurations.

When NGINX processes HTTP/2 client connections and forwards requests to backend services, a specially crafted request can trigger the issue if the NGINX worker process is running low on available worker connections.

The affected worker may continue referencing a connection that has already been released.

The potential consequences include denial of service and remote code execution under additional conditions involving ASLR.

This is particularly concerning for organizations operating heavily utilized NGINX reverse proxies, where worker connection availability is an important operational consideration.

CVE-2026-123007 – HTTP/2 Integer Underflow

The second vulnerability originates in NGINX’s HTTP/2 flow-control implementation.

HTTP/2 uses flow control to regulate the amount of data transmitted between communicating endpoints. NGINX maintains a send-window value to manage this process.

According to the disclosed research, incorrect flow-control accounting can cause the send-window value to exceed its permitted threshold.

Subsequent arithmetic operations can result in an integer underflow, leading to an invalid memory-length calculation and a heap buffer overflow.

This can cause an NGINX worker to crash or corrupt memory, with potential remote code execution under the reported exploitation conditions.

One important aspect of this vulnerability is its reach.

Wiz demonstrated code execution against an NGINX server serving ordinary static content. The demonstration did not require a backend application or third-party modules.

The vulnerability affects NGINX Open Source versions dating back to 1.9.5, released in 2015.

That means a memory-safety weakness remained in a widely deployed web infrastructure component for more than a decade before its disclosure.

Why these vulnerabilities matter

Both vulnerabilities affect NGINX itself rather than the applications hosted behind it.

This distinction is important.

Organizations often concentrate their security assessments on application vulnerabilities, API security and backend infrastructure. However, the reverse proxy is also part of the application delivery architecture and deserves the same level of security attention.

A successful compromise of an NGINX server could potentially allow an attacker to:

  • Execute arbitrary code on the affected server.
  • Access sensitive information available to the NGINX process.
  • Manipulate website content and application traffic.
  • Disrupt services hosted behind the affected deployment.
  • Potentially use the compromised server as a stepping stone for further attacks.

Both vulnerabilities require HTTP/2 to be enabled on the affected NGINX server. However, their additional exploitation conditions differ.

The important distinction is that CVE-2026-123633 involves a specific reverse-proxy and worker-connection condition, whereas CVE-2026-123007 can also affect deployments serving static content directly.

Affected versions and security updates

F5 has released security updates addressing both vulnerabilities.

For NGINX Open Source, the fixed releases are:

  • Stable branch: 1.30.7
  • Mainline branch: 1.31.9

For NGINX Plus, the required fixed releases differ between the two CVEs.

For CVE-2026-123633, the listed fixed releases include 37.0.7.1 and 37.2.0.1.

For CVE-2026-123007, the listed fixed releases include 37.0.8.1 and 37.3.0.1.

Organizations should consult the respective F5 advisories and select a release that addresses both vulnerabilities. Older supported release branches may require an upgrade to a fixed branch.

What should security teams do?

From a Vulnerability Management perspective, this requires more than simply checking the NGINX version.

Asset discovery

Identify NGINX deployments across on-premises infrastructure, cloud environments, containers and Kubernetes clusters. Include reverse proxies, internet-facing web servers and internal application gateways.

Configuration assessment

Determine whether HTTP/2 is enabled on client-facing connections. Identify deployments forwarding requests to backend services and assess worker connection configurations.

Exposure prioritization

Prioritize internet-facing NGINX deployments, particularly those supporting business-critical applications and multiple backend services.

Patch management

Apply the relevant F5 security updates. Validate the actual running binary and ensure that old worker processes have exited following the upgrade.

Temporary mitigation

Where immediate patching is not possible, temporarily disable HTTP/2 on affected NGINX listeners. This reduces exposure to these two vulnerabilities but does not replace the security update.

Threat hunting

Investigate unexpected NGINX worker crashes, unusual process activity and suspicious commands launched by NGINX workers. These are general investigation indicators, not confirmed indicators of exploitation.

The bigger picture

The most important takeaway from these disclosures is not simply the presence of two critical vulnerabilities in NGINX.

It is the location of the weaknesses.

NGINX frequently sits at the front door of enterprise applications. It handles incoming connections, terminates traffic and routes requests to services that may otherwise have strong application-level security controls.

A vulnerability at this layer can undermine the security assumptions of multiple applications simultaneously.

A secure application behind a vulnerable reverse proxy is not necessarily a secure application.

Comments

No comments yet. Why don’t you start the discussion?

    Leave a Reply

    This site uses Akismet to reduce spam. Learn how your comment data is processed.