Apache mod_http2 RCE (CVE-2026-119618): check HTTP/2 and patch
CVE-2026-119618 is a race condition in Apache httpd’s mod_http2 that corrupts heap memory and, according to Apache, can lead to remote code execution. Apache rates it critical. It affects every release from 2.4.55 through 2.4.70, but only on servers where HTTP/2 is enabled; a server that speaks HTTP/1.1 only is not exposed. The fix is 2.4.71, released on 11 November 2026, or a distribution package that lists the CVE. Turning HTTP/2 off is the stopgap. A public proof of concept crashes Apache without running code; Red Hat is quoted as saying public research has shown a reliable exploit against a default configuration. No exploitation in the wild had been reported by 13 November. For a WordPress host this sits below WordPress entirely: no plugin sees the request and no WordPress update fixes it.
Prepared by G. Schad, WordPress and Linux security specialist
First published in the weekly briefing of 16 November 2026 and moved here so the checks and the stopgap have a page of their own. The modified date changes when the exploitation status does.
Affected
Apache httpd 2.4.55 through 2.4.70, and mod_h2 2.0.8 through 2.0.42, with h2 or h2c in a Protocols directive
Fixed in
2.4.71 (11 November 2026); distribution packages backport it, so check the changelog for the CVE
Exploitation
Public PoC reproduces the crash; Red Hat reportedly says a reliable exploit has been shown. No reports in the wild, not in CISA’s KEV as of 13 November
Stopgap
Protocols http/1.1 on every server and virtual host, then a full restart
Apache httpd · CVE-2026-119618
Apache 2.4.71: a critical mod_http2 flaw in every release since 2.4.55, including 2.4.70
Apache released httpd 2.4.71 on 11 November with three security fixes. The first is the most serious Apache flaw this year for anyone hosting WordPress on it. CVE-2026-119618 is a race in the assign_headers() function of mod_http2. Apache describes it as concurrent execution using a shared resource without proper synchronisation that “can lead to RCE”, and rates it critical. It affects 2.4.55 through 2.4.70, and the standalone mod_h2 module from 2.0.8 through 2.0.42. Two teams found it independently: Anthropic’s OSS Scanner reported it on 26 October, and Google Cloud Vulnerability Research reported it on 3 November. The fix is commit 824beab. Published patch-diff analyses describe it as a one-line change that writes a synthesised header into the request’s own header table instead of a table shared by the connection’s streams.
The condition that matters is HTTP/2. Red Hat puts it plainly: only servers with HTTP/2 enabled are affected, and servers that serve only HTTP/1.1 are not. Apache’s own default for the Protocols directive is http/1.1, but many hosting stacks turn HTTP/2 on for TLS sites because browsers prefer it. For each server, the question is whether any Protocols line, server-wide or in a virtual host, lists h2 or h2c.
The scores differ depending on who assigned them. Apache publishes no number, only “critical”. The CVE record from Apache’s CNA scores it 8.1, with high attack complexity. Red Hat scores it 9.8, but rates its own RHEL 9 and 10 packages Important. NVD had not analysed it at the time of writing. The gap comes from the race: an attacker has to win it, which is why the CNA rates the complexity as high.
Exploitation status, from the sources we could check, as of 13 November. Wiz wrote on 11 November that no exploit was available and nothing had been seen in the wild. On 12 November, SecurityWeek reported a public proof of concept on GitHub that reproduces the memory corruption and crashes Apache, but does not show code execution. Red Hat, as quoted by Rescana on 13 November, says public research has shown a reliable exploit against a default-configured server with mod_http2 enabled, without naming it. The flaw was not in CISA’s KEV catalog on 13 November. Wiz’s telemetry finds resources running an affected version in 64% of cloud environments, and publicly exposed ones in 51%, without accounting for whether HTTP/2 is on. Patch it this week; don’t leave it for the next maintenance window.
The same release fixes two smaller flaws. CVE-2026-106023, rated important, is a NULL pointer dereference in mod_md during request processing that crashes a child process. It affects 2.4.30 through 2.4.70 and matters on servers that use mod_md for Let’s Encrypt certificates. CVE-2026-109163, rated low, is a one-byte stack overwrite in mod_vhost_alias in 2.4.69 and 2.4.70, a follow-up to CVE-2026-63292. Neither is the reason to update today; the HTTP/2 flaw is.
One source of confusion: CVE-2026-23918, the HTTP/2 double free on early reset that 2.4.67 fixed in May, is a different bug, and it affected 2.4.66 only. A server patched for that one is still exposed to this one.
Figure 1 · At a glance
If HTTP/2 is on and Apache is 2.4.55 to 2.4.70, patch it this week.
Question
Answer
Affected
Apache httpd 2.4.55 through 2.4.70 with h2 or h2c enabled
Not affected
Servers that serve HTTP/1.1 only; RHEL 8 and earlier
Fixed in
2.4.71, 11 November 2026, or a package listing the CVE
Exploitation
Public PoC crashes Apache; reliable exploit reported by Red Hat; none in the wild by 13 Nov
If you cannot patch
Protocols http/1.1 everywhere, configtest, full restart
The points that decide what to do about the Apache httpd mod_http2 flaw CVE-2026-119618, as stated in the sections below.
Is HTTP/2 on, is the fix in, and what to do if you cannot patch today
Run these on every server where Apache faces the internet or a proxy, including containers. A version string alone can mislead. Distributions backport the fix under an older version number. The mod_http2 version string is no guide either: the fix went into httpd’s own copy of the module, no new standalone mod_h2 release had been tagged by 13 November according to Rescana, and one administrator reported that the official 2.4.71 Docker image still reports mod_http2 2.0.42.
01
Which Apache is running, and is mod_http2 loaded
The version from the binary, and the module list from the running configuration:
A 2 from the first command means h2 over TLS; a 2 from the second means cleartext h2c. A CDN that speaks HTTP/2 to visitors does not protect an origin that also answers HTTP/2 directly.
04
Whether the installed package carries the fix
Read the changelog, since a backported package keeps the old version number:
On RHEL 9 and 10 the fix ships in the mod_http2 package, not httpd; Red Hat says RHEL 8 and earlier are not affected. On cPanel, compare the ea-apache24 packages against the EasyApache 4 change log; don’t assume the nightly update has already run.
05
Restart Apache fully
A full restart is the reliable way to be sure no process is still running the old code. After the update or the configuration change:
shell
$ apachectl configtest &&systemctl restart httpd
Use apache2 on Debian and Ubuntu, and /scripts/restartsrv_httpd on cPanel. Rebuild and redeploy any container image that bundles httpd: pulling the new base image does nothing for containers that are already running.
06
If you cannot patch today, turn HTTP/2 off
Change every Protocols line that lists h2 or h2c to HTTP/1.1 only, check the configuration and restart:
shell
$ Protocols http/1.1
Sites keep working over HTTP/1.1 with TLS, and browsers fall back without errors. Put HTTP/2 back after the update. This is the mitigation Red Hat and Wiz give; Apache publishes no workaround and says to upgrade.
07
Look back to 11 November
A failed attempt leaves repeated child crashes in the error log. A successful one can leave a shell started by a web-server process:
shell
$ grep-hE"exit signal (Segmentation fault|Aborted)" /var/log/httpd/error_log /var/log/apache2/error.log /usr/local/apache/logs/error_log 2>/dev/null|tail -20$ for p in $(pgrep-x httpd;pgrep-x apache2);dops--ppid"$p"-o pid,user,lstart,args --no-headers;done|grep-vE"httpd|apache2|php"
The second command lists the direct children of every Apache process, minus Apache and PHP themselves. A shell, a downloader or another interpreter in that list is not normal, and is a reason to open a case.
Is my WordPress site affected by the Apache HTTP/2 flaw, CVE-2026-119618?
Only if the server runs Apache httpd 2.4.55 to 2.4.70 with HTTP/2 enabled. WordPress itself is not involved: the flaw is reached before any PHP runs. If your sites are served by nginx or LiteSpeed rather than Apache, it does not apply. If nginx sits in front of Apache, the backend is usually reached over HTTP/1.1, but check that Apache does not also answer HTTP/2 directly. On managed hosting, ask the host whether its Apache is on 2.4.71 or a fixed package.
Will turning off HTTP/2 slow my site down?
A little, for pages with many assets, and only until you patch. Browsers fall back to HTTP/1.1 over the same TLS connection without errors. Compared with a critical remote flaw, a slower first load for a day or two is the cheaper option. Turn HTTP/2 back on after the update.
I updated Apache to 2.4.70 last week. Am I covered?
No. 2.4.70, released on 6 November, is the last affected release. Go to 2.4.71, or to your distribution’s package that lists CVE-2026-119618 in its changelog, and restart Apache.
Choose the response path
Apache runs HTTP/2 on more servers than anyone listed: choose the response path.
The servers under the sites
Get Apache, HTTP/2 and the rest of the web stack checked on every host.
Best when several servers need the Apache version, the protocols per virtual host and the backported packages verified from the running process, the look-back since 11 November done once, and a written result per host.
Get the finding classified and the site restored to a trusted state.
Best when a file-upload plugin left PHP in uploads, wp-config.php went missing, the tampered Quick Page/Post Redirect build turned up, or an account changed hands, and you want every layer removed in one pass.