
Coordinated Disclosure and Critical Scope
Rapid7 and F5 have jointly disclosed CVE-2026-123007, a critical integer underflow vulnerability in nginx HTTP/2 processing that can allow unauthenticated remote code execution. Published on November 18, 2026 by researcher Stephen Fewer in a Rapid7 Intelligence advisory, the flaw carries a CVSSv4 base score of 9.2 and is classified under CWE-191 for integer underflow. Although the advisory confirms that upstream fixes are available, the defect affects core HTTP/2 response handling under specific conditions and exposes unpatched servers to memory corruption within the nginx worker process.
According to Rapid7, the bug stems from HTTP/2 flow-control accounting errors that let a remote, unauthenticated client corrupt worker memory without needing prior credentials or local access. Depending on how an attacker sequences requests and connection state, the resulting corruption can leak sensitive in-memory data, disrupt availability through service interruption, or achieve arbitrary code execution under the permissions of the nginx worker process. Because nginx underpins vast segments of open-source Linux web infrastructure, reverse proxies, and TLS termination tiers, any remote memory-corruption primitive in its HTTP/2 stack warrants immediate patch prioritization.
HTTP/2 Flow-Control Mechanics and Proof-of-Concept Metrics
At the protocol level, CVE-2026-123007 occurs when nginx miscalculates the volume of response data an HTTP/2 client has permitted the server to send. Rapid7 explains that a remote client can exploit this accounting mistake by cancelling queued response data and then forcing nginx to apply an invalid length parameter when copying a subsequent response. That invalid length drives out-of-bounds memory operations inside the worker process, bridging flow-control state manipulation into full worker-context memory corruption.
To validate exploitability beyond theoretical crash analysis, Rapid7 built a proof-of-concept targeting a vulnerable linux/amd64 Docker image running nginx version 1.31.8 with TLS enabled and the http2 on directive active. In a demonstration video documented as Figure 1 and sped up to 8x playback for brevity, Rapid7 achieved unauthenticated remote code execution in two and a half minutes. Reaching reliable exploitation is notably bandwidth-intensive, requiring the client and server to exchange approximately 128 GB of HTTP/2 network traffic. Rapid7 estimates that transmitting this required 128 GB payload takes roughly two minutes over a 10 Gbit/s network link and about seventeen minutes over a 1 Gbit/s link.
Operational Impact for Linux Servers and DevOps Workflows
For systems administrators, platform engineers, and DevOps teams maintaining open-source nginx deployments, the practical risk depends heavily on listener configuration and network capacity. Any public-facing nginx instance serving HTTP/2 over TLS on high-bandwidth cloud virtual machines, bare-metal Linux hosts, or container ingress controllers sits squarely within the exposure profile outlined by Rapid7. Even where full remote code execution proves difficult to time reliably, the same underlying memory-corruption flaw can trigger worker crashes or disclose arbitrary memory contents from shared worker buffers, potentially exposing session tokens, headers, or upstream proxy payloads.
Integrating upstream remediation into production pipelines requires coordinating container base-image rebuilds, Linux distribution package updates, and custom-compiled nginx binaries. Teams running nginx 1.31.8 or neighboring unpatched releases should consult the linked Rapid7 advisory and F5 security bulletins to verify exact patched version numbers and interim configuration guidance before deploying updates. In the meantime, security operations teams can pair patching efforts with network telemetry reviews, watching for unusually long-lived TLS-wrapped HTTP/2 sessions transferring tens of gigabytes of data to a single worker, and restricting unauthenticated access to endpoints capable of queueing large responses where feasible.
Exploitation Constraints, Patch Verification, and Reporting Caveats
Despite its 9.2 CVSSv4 severity, CVE-2026-123007 faces concrete physical and protocol constraints on the open Internet. Rapid7 notes that available bandwidth and nginx connection timeouts sharply narrow viable targets. Because nginx enforces a default one-hour connection lifetime, an attacker operating over a link slower than roughly 300 Mbit/s cannot transfer the necessary 128 GB of HTTP/2 traffic before the session times out, preventing successful exploitation. High-speed data-center networks, cloud-to-cloud paths, and internal lateral connections remain far more exposed than congested residential or low-throughput edges.
Readers should also note clear evidentiary boundaries in this summary. Rapid7's disclosure indicates that additional HTTP/2 connection conditions affect exploitability, and full remediation steps and disclosure timelines appear in the complete advisory beyond the summary excerpt reviewed here. No independent reproduction, patch diffing, or custom throughput testing was conducted for this article; all version details, timing estimates, and proof-of-concept metrics are attributed directly to Rapid7 Intelligence and F5. Operators should treat the 128 GB threshold and timing figures as reference measurements from Rapid7's linux/amd64 nginx 1.31.8 test case rather than universal guarantees across every architecture or distribution build.
Automated Vertex AI (gemini-4-hydrogen) primary-source report; reviewed 2026-11-19.



