Critical unauthenticated RCE flaws found in NGINX HTTP/2 handling
Two critical remote code execution vulnerabilities, CVE-2026-123633 and CVE-2026-123007, affect NGINX servers with HTTP/2 enabled. F5 has released patches for all supported versions.
Security researchers have identified two critical remote code execution vulnerabilities in NGINX, the world’s most widely used open-source web server. Disclosed on November 18, 2026, by Wiz Research alongside independent teams from Ikotas Labs, Metnir, Rapid7, and Xint, these flaws allow unauthenticated attackers to take full control of affected servers. The vulnerabilities, tracked as CVE-2026-123633 and CVE-2026-123007, specifically impact NGINX instances that have HTTP/2 protocol support enabled.
What happened
The discovery centers on memory-safety issues within the NGINX core itself, meaning that even applications with secure backend code are vulnerable if they sit behind an unpatched NGINX instance. Wiz Research utilized AI-assisted code scanning tools to uncover these long-standing bugs. CVE-2026-123007 involves an integer underflow in HTTP/2 flow-control accounting and affects versions dating back to NGINX 1.9.5, released in 2015. This means the flaw existed in the codebase for over a decade before detection. The second vulnerability, CVE-2026-123633, is a use-after-free error in HTTP/2 connection handling that has been present for nearly six years.
Both vulnerabilities carry a CVSS 4.0 score of 9.2, indicating critical severity. Successful exploitation allows an attacker to execute arbitrary code on the server without any authentication. In testing, researchers demonstrated that CVE-2026-123007 could be exploited on servers serving only static files, requiring no backend application or third-party modules. Because NGINX often acts as a reverse proxy for multiple services, a single compromised server can expose numerous websites and APIs to data theft, tampering, or service disruption.
Key details
- Vulnerability IDs: CVE-2026-123633 (use-after-free) and CVE-2026-123007 (integer underflow).
- Severity: Both are rated Critical with a CVSS 4.0 score of 9.2.
- Prerequisite: HTTP/2 must be enabled on the NGINX server for exploitation.
- Affected Versions: CVE-2026-123007 affects NGINX 1.9.5 through 1.31.8; CVE-2026-123633 affects 1.19.7 through 1.31.8.
- Fixed Releases: NGINX OSS 1.30.7 (stable) and 1.31.9 (mainline); NGINX Plus 37.0.7.1, 37.2.0.1, 37.0.8.1, and 37.3.0.1.
- Mitigation: If immediate patching is not possible, administrators should disable HTTP/2 by setting
http2 off;in server blocks.
Background
NGINX serves as the front door for more than 225 million websites, functioning as a web server, reverse proxy, and load balancer. It accepts incoming traffic and routes it to the appropriate application servers. The vulnerabilities discovered here are memory-safety flaws, which occur when software incorrectly manages memory allocation and deallocation. In a use-after-free scenario, the program continues to use a pointer to memory after it has been freed, potentially allowing an attacker to inject malicious code. An integer underflow happens when a calculation results in a number smaller than the minimum limit, leading to unexpected behavior such as invalid memory copies.
Historically, finding such deep-seated memory errors required extensive manual auditing. However, this disclosure highlights a shift in vulnerability research. Wiz Research employed AI models, specifically Gemini 4 Carbon and ChatGPT 5.6 Sol, to validate the impact of these flaws. The ability of AI to trace complex interactions between memory operations and code paths is lowering the barrier to discovering longstanding vulnerabilities. This accelerates both defensive remediation and the potential for offensive exploitation, shrinking the window between discovery and active threat.
Why it matters
For teams managing self-hosted infrastructure, this incident underscores that security is not solely about application code. Even if your development team writes perfect Python, Go, or Node.js code, a vulnerability in the underlying web server like NGINX can bypass those defenses entirely. Since NGINX is often shared across multiple internal tools and customer-facing sites, a single unpatched instance creates a broad attack surface. The fact that one of these flaws existed for eleven years suggests that many common configurations may harbor similar dormant risks.
The reliance on HTTP/2 is another critical factor. HTTP/2 is widely enabled to improve performance and latency, making it a default choice for many modern deployments. Disabling it is a viable emergency mitigation, but it may degrade user experience or break compatibility with certain clients. Therefore, rapid patching is essential. The coordinated disclosure process involved multiple independent teams finding the same issues, indicating that these flaws were likely nearing public exploitation. Delaying updates increases the risk of encountering automated attacks scanning for these specific CVEs.
What you can do
- Inventory your NGINX instances: Identify all servers running NGINX versions between 1.9.5 and 1.31.8, or 1.19.7 and 1.31.8 for the second CVE.
- Check HTTP/2 status: Verify if HTTP/2 is enabled in your configuration files. If it is disabled, your systems are not vulnerable to these specific flaws.
- Apply patches immediately: Upgrade to NGINX OSS 1.30.7 or 1.31.9, or the corresponding NGINX Plus releases listed in the advisory.
- Implement temporary mitigation: If you cannot patch right away, disable HTTP/2 by adding
http2 off;to your server blocks and removing thehttp2parameter from listen directives. - Test and reload: After changing configurations, test application compatibility, reload NGINX, and ensure old worker processes have exited.
- Monitor for exploitation: Watch logs for unusual HTTP/2 traffic patterns or unexpected process behavior while waiting for patch deployment.



