A critical flaw in Apache HTTP Server’s HTTP/2 module is now being exploited in real attacks, according to a fresh advisory from Singapore’s Cyber Security Agency (CSA). The warning lands roughly a week after Apache quietly shipped a fix, underscoring how quickly a patched vulnerability can turn into an active threat once technical details and proof-of-concept code begin circulating.
A Race Condition Hiding in Plain Sight
The vulnerability, tracked as CVE-2026-119618, lives inside mod_http2, the module that lets Apache serve traffic over the newer HTTP/2 protocol. At the center of the bug is a function called assign_headers(), which is supposed to manage how multiple execution threads share access to the same piece of memory while processing incoming requests.
According to Apache’s own advisory, that sharing isn’t properly synchronized. Two threads can end up racing to use the same resource at the same time, and depending on which one “wins,” the result can corrupt the server’s heap memory. In the worst case, an attacker who can trigger that race condition reliably enough can turn simple memory corruption into full remote code execution — meaning they could run their own commands with whatever privileges the web server process holds.
Crucially, no authentication is required. An attacker only needs the ability to send specially crafted HTTP/2 requests to a vulnerable server.
How Serious Is It, and Who’s Exposed?
Apache itself rates the bug as critical. Singapore’s CSA assigned it a CVSS 3.1 score of 8.1, placing it firmly in high-severity territory. The affected range is broad: Apache’s security advisory lists versions 2.4.55 through 2.4.70 as vulnerable whenever mod_http2 is active, along with standalone mod_h2 builds from 2.0.8 through 2.0.42.
Cloud security firm Wiz, which analyzed the issue shortly after the patch shipped on November 11, found that a striking share of real-world deployments remain exposed. In its scan of cloud environments, roughly 64% had resources running an affected Apache version, and about 51% had instances reachable directly from the internet. Wiz was careful to note that these figures only describe where HTTP/2 traffic could theoretically be at risk — they don’t confirm that HTTP/2 was actually enabled on every one of those servers, nor do they indicate how many have actually been breached.
That nuance matters for triage. A server running an affected version but configured to speak only HTTP/1.1 isn’t exposed to this particular flaw. Administrators are better off checking which protocols are actually active on their listeners rather than relying on version numbers alone.
Credit Where It’s Due
Apache credited two independent discoveries behind the fix: an automated scan from Anthropic’s open-source vulnerability scanner flagged the issue on October 26, and security researchers at Google Cloud reported it separately just over a week later, on November 3. The overlap between an automated tool and manual human research is a reminder that AI-assisted scanning is increasingly finding its way into serious vulnerability discovery pipelines — not just bug bounty submissions.
It’s also worth noting this isn’t the first HTTP/2-related remote execution bug Apache has had to deal with this year; a separate flaw in the same subsystem was fixed back in May. According to the advisory, that earlier fix does nothing to protect against this new issue, since the two bugs have different root causes.
What Administrators Should Do Right Now
The clearest fix is to update to Apache HTTP Server 2.4.71 or later, or to install whatever package your Linux distribution or cloud vendor has confirmed contains the backport. For teams that can’t patch immediately, Apache and researchers suggest a temporary mitigation:
- Set
Protocols http/1.1at the server level to disable HTTP/2 entirely. - Double-check that individual virtual hosts don’t override this and still allow
h2orh2c. - Reload the service and verify, using a protocol-aware test, that the server no longer negotiates HTTP/2 connections.
- Prioritize internet-facing servers first, but don’t ignore internal-only deployments — lateral movement after an initial compromise elsewhere in the network remains a realistic scenario.
For any server that was running an affected, HTTP/2-enabled configuration before the patch was applied, security teams should treat the period before the update as a potential exposure window and review logs for unusual process activity, unexpected crashes, or signs of lateral movement. Installing the fix closes the hole going forward, but it offers no retroactive assurance that the server wasn’t already touched.
With active exploitation now confirmed and a public proof-of-concept reportedly available, the gap between “patch released” and “patch urgent” has effectively closed for anyone still running vulnerable, HTTP/2-enabled Apache instances.
Leave a Reply
You must be logged in to post a comment.