
Two Critical NGINX RCE Flaws Found in HTTP/2 Handling
Researchers uncovered two critical unauthenticated remote code execution vulnerabilities in NGINX, tracked as CVE-2026-123633 and CVE-2026-123007, both rated 9.2 and requiring HTTP/2.
Wiz Research, Ikotas Labs, Metnir, Rapid7, and Xint independently uncovered two critical unauthenticated remote code execution vulnerabilities in NGINX, tracked as CVE-2026-123633 and CVE-2026-123007. Both are rated Critical with a CVSS 4.0 score of 9.2, require HTTP/2 to be enabled, and were reported to F5, which released fixes on November 18, 2026. Successful exploitation could let attackers execute code, steal data, tamper with websites, or disrupt services on affected servers.
Impact and affected versions
NGINX powers more than 225 million websites and often acts as a shared entry point for several applications. Because the flaws are in NGINX itself, websites and APIs can be exposed even when their own application code is secure. CVE-2026-123633 is a use-after-free in HTTP/2 connection handling and affects reverse-proxy processing when available connection slots run low. CVE-2026-123007 is an integer underflow in HTTP/2 flow-control accounting and can affect NGINX serving content directly, including ordinary static files over HTTPS on 64-bit systems. Wiz demonstrated remote code execution for CVE-2026-123007 without a backend application or third-party module. Typical mitigations such as ASLR do not protect against these issues because attackers can use either flaw for memory read access before code execution. The integer underflow dates back to NGINX 1.9.5 in 2015, while the use-after-free had been present for nearly six years.
Mitigation and fixed releases
Organizations should identify affected deployments with HTTP/2 enabled and upgrade as soon as possible. Fixed releases include NGINX OSS 1.30.7 stable and 1.31.9 mainline for both CVEs; NGINX Plus fixes are 37.0.7.1 and 37.2.0.1 for CVE-2026-123633, and 37.0.8.1 and 37.3.0.1 for CVE-2026-123007. If immediate upgrading is not possible, temporarily disable HTTP/2 by setting http2 off in affected server blocks on NGINX 1.25.1 or later and removing the http2 parameter from listen directives. Afterward, test application compatibility, reload NGINX, confirm old workers have exited, and re-enable HTTP/2 after upgrading. Unexpected NGINX worker crashes or commands launched by NGINX are general warning signs, not confirmed indicators of these vulnerabilities.
How the flaws work
In CVE-2026-123633, an HTTP/2 connection can remain on NGINX's reusable connection list after a request stream starts. If connection slots run low while forwarding the request, NGINX may close that connection and free its memory, but request processing can continue using the freed memory. In CVE-2026-123007, returning credit for canceled streams can push the send_window counter above HTTP/2's maximum. A later bounds check underflows because of unsigned integer arithmetic, allowing further increases. The invalid counter is later converted to a signed integer, producing a negative length and an out-of-bounds memory copy during TLS sending. Wiz confirmed remote code execution with an ASLR bypass on official NGINX 1.31.7 and 1.31.8 builds using HTTP/2 over TLS on 64-bit systems.
Exploitation and detection
Wiz built proof-of-concept exploits but is not releasing them. Exploitation is difficult and requires memory disclosure, heap grooming, and handling different NGINX versions, builds, and configurations. Attackers may target packaged products where many customers share the same build. Any exploit requires direct communication with the affected NGINX server, and CDNs or WAFs that terminate HTTP/2 may mitigate parts of the attack. CVE-2026-123007 requires sending many WINDOW_UPDATE records, so detection rules for long-standing HTTP/2 connections sending large amounts of data may hint at an attack. CVE-2026-123633 is more likely to be exploited when NGINX is configured as a proxy.
Sources: wiz.io
SiTech — AI-powered web development
We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.