A critical flaw in Apache HTTP Server, one of the most widely used pieces of web server software in the world, could let an attacker on the internet take control of a company's web server without logging in. Apache released a fix on November 11. Only servers with the newer HTTP/2 protocol switched on are at risk, but that setting is common, and some Linux distributions and appliances turn it on by default. Cloud security firm Wiz found affected versions in 64% of the cloud environments it examined, with 51% exposing them directly to the internet.
There are no confirmed attacks yet, but the picture is unsettled. Red Hat says public research has already shown a reliable way to exploit the flaw; Wiz says no exploit is available. Organisations running Apache should not wait for that question to be settled. Patch internet-facing servers first, and if that cannot happen quickly, turn HTTP/2 off until it can. Note that for WordPress and similar sites, updating the website software will not help; the flaw is in the web server underneath.
The vulnerability is tracked as CVE-2026-119618. It sits in mod_http2, the module that handles HTTP/2 traffic, and affects Apache HTTP Server 2.4.55 through 2.4.70, as well as the standalone mod_h2 2.0.8 through 2.0.42. Apache rates it Critical, and Red Hat gives it a CVSS score of 9.8. It is fixed in Apache HTTP Server 2.4.71.
Who is affected
Exposure depends on two things: the version you run and whether HTTP/2 is enabled. Version alone does not make a server exploitable.
Potentially affected:
- Apache HTTP Server 2.4.55 through 2.4.70 with
mod_http2loaded and HTTP/2 enabled - Standalone
mod_h2builds from GitHub, 2.0.8 through 2.0.42 - Servers that list
h2(HTTP/2 over TLS) orh2c(HTTP/2 over plain TCP) in theProtocolsdirective, either server-wide or in any single virtual host
Not affected:
- Servers configured to serve HTTP/1.1 only
- Apache versions earlier than 2.4.55
- Red Hat Enterprise Linux 6, 7, and 8, according to Red Hat
Two groups deserve particular attention. First, servers that were patched in May for CVE-2026-23918, an earlier mod_http2 flaw fixed in 2.4.67, are still exposed: 2.4.67 through 2.4.70 are all vulnerable to this new issue. Second, the standalone mod_h2 project's repository was archived on June 9, and The Hacker News reports there is no fixed standalone release. Servers using it should move to the mod_http2 shipped with Apache 2.4.71 or disable HTTP/2.
| Product | Affected versions | Fixed version |
|---|---|---|
| Apache HTTP Server | 2.4.55 to 2.4.70 | 2.4.71 |
Standalone mod_h2 (GitHub) |
2.0.8 to 2.0.42 | No standalone fix reported; repository archived |
How the flaw works
HTTP/2 lets a browser send many requests over a single connection at the same time, each as its own "stream". Apache handles those streams in parallel using multiple worker threads, which is where this flaw lives.
CVE-2026-119618 is a race condition in the assign_headers() function of mod_http2. Threads handling concurrent HTTP/2 requests share resources without proper locking, so two code paths can touch the same memory at the same time. According to Praetorian's analysis, a worker thread can end up writing to memory that belongs to another thread through a shared allocator, corrupting the server's heap. Heap corruption of this kind is the building block for remote code execution: an attacker who can shape it reliably may be able to run commands with the privileges of the Apache server process.
Apache's advisory gives little technical detail. Race conditions are often harder to exploit reliably than straightforward memory bugs, because the attacker has to win a timing contest, and that is reflected in how vendors are scoring this issue.
Why Red Hat rates its own packages lower
Red Hat assigns the CVE a 9.8 but rates its own RHEL 9 and RHEL 10 packages Important rather than Critical. The reason is memory allocation: on RHEL 9 and 10, each HTTP/2 stream gets its own allocator, which Red Hat says makes exploitation substantially harder. Other builds may not have that protection, so the lower RHEL rating should not be read as a lower rating for Apache in general.
Severity
| Source | Rating | Notes |
|---|---|---|
| Apache | Critical | No CVSS score published; the CVE record was still marked reserved at the time of reporting |
| Red Hat (CVE) | 9.8 (CVSS) | Remote, unauthenticated; vector not yet published in public reporting |
| Red Hat (RHEL 9 and 10 packages) | Important | Separate per-stream allocators make exploitation substantially harder |
The attack is network-based and needs no account and no user interaction. The main prerequisite is that HTTP/2 is enabled. Because the CVE record was still reserved when this was written, scanners that rely on NVD scores may not yet show it with a severity, or at all.
Exploitation status
As of November 12, there are no confirmed reports of exploitation in the wild. Beyond that, the sources disagree.
- Red Hat says that public research has demonstrated a reliable exploit against a server with a default configuration and
mod_http2enabled, and reports remote code execution. It has not named or linked that research. - Wiz said in its November 11 blog post that no exploit was available and there were no reports of exploitation in the wild.
- A public proof of concept on GitHub reproduces the memory corruption and crashes the server process, but does not demonstrate code execution.
- Praetorian is building its own exploit to test whether code execution is achievable, and had not validated it at the time of its November 12 post.
In practical terms: crashing Apache through this flaw is already public, which means denial of service is a realistic near-term risk even if code execution proves difficult. Red Hat's statement suggests that someone, somewhere, believes code execution is achievable against default setups. Treat the more cautious "no exploit" view as a statement about what has been published, not about what attackers have.
We have not tested any of the public code, and we recommend against running unvetted exploit code outside an isolated lab.
What you should do
1. Find out where you are exposed
Version and configuration both matter, so check both.
- Confirm the module is loaded: run
apachectl -M | grep http2, or look forLoadModule http2_modulein your configuration. - Confirm HTTP/2 is enabled: search your configuration for
Protocolslines containingh2orh2c. Check the global setting and every virtual host; one forgotten virtual host is enough. - Check the version, carefully: Linux distributions often backport fixes without changing the version number, so a patched server may still report a 2.4.x version below 2.4.71. Use your vendor's advisory and package version, not the Apache version string alone.
- Prioritise: internet-facing servers first, then reverse proxies, and do not overlook Apache embedded in appliances or container images, which are easy to miss in an inventory.
2. Patch, then restart
Upgrade to Apache HTTP Server 2.4.71, or install your vendor's fixed package. Restart httpd afterwards; until the running processes reload, the old module stays in memory.
Distribution status as of November 12:
| Platform | Status |
|---|---|
| RHEL 9 | Fixed, mod_http2-2.0.26-7.el9_9.1 (rated Important) |
| RHEL 10 | Fixed, mod_http2-2.0.29-5.el10_3.1 (rated Important) |
| RHEL 9.4 SAP, 9.6, 9.8, 10.0, 10.2 EUS | Fixed, across several advisories |
| RHEL 6, 7, 8 | Not affected |
| Red Hat JBoss Core Services, Hardened Images | Affected, no fix listed |
| Debian sid | Fixed, apache2 2.4.71-1 |
| Debian bookworm, trixie, forky | Vulnerable, no fix listed |
| Amazon Linux 2, 2023, 2027 Preview | Fix pending |
| Ubuntu | No status page published at the time of reporting |
If your platform has no fix yet, apply the workaround below and keep checking.
3. If you cannot patch today, turn HTTP/2 off
Apache has published no official workaround. The interim step recommended by Wiz, Praetorian, and others is to stop serving HTTP/2:
- Remove
h2andh2cfrom everyProtocolsdirective, leavingProtocols http/1.1, or unloadmod_http2entirely. - Apply this at the server level and in each virtual host.
- This also covers servers running the standalone
mod_h2.
Browsers and most clients fall back to HTTP/1.1 automatically, but test first where an application or API depends on HTTP/2, since performance and some clients may be affected. This reduces exposure; it does not replace the patch.
4. Do not rely on application updates
For WordPress, Drupal, or any other application running on Apache, updating the application or its plugins does not fix this. Wordfence has warned WordPress site owners about exactly this. If you use managed or shared hosting, ask your provider whether its Apache servers are patched or have HTTP/2 disabled.
5. Watch for signs of trouble
No indicators of compromise have been published. Until they are:
- Monitor Apache error logs for unusual crashes, segmentation faults, or repeated worker process restarts. Since the public proof of concept crashes the server, this is the most likely early signal of probing.
- Watch for the Apache process (often
httpd,apache2, or thewww-data/apacheuser) spawning shells or utilities such assh,bash,curl, orwget, which a web server should almost never do. - Look for unexpected outbound connections from web servers, and for new files written to temporary directories by the web server user.
6. Patch the rest of the 2.4.71 release
Apache 2.4.71 also fixes two other issues:
- CVE-2026-106023 (Important): a NULL pointer error in
mod_mdthat can crash a child process while handling a request. Affects 2.4.30 through 2.4.70, a much wider range than the HTTP/2 flaw. - CVE-2026-109163 (Low): a one-byte stack overflow in
mod_vhost_alias, affecting only 2.4.69 and 2.4.70.
Upgrading to 2.4.71 handles all three.
A note on the pattern
This is the latest in a run of serious mod_http2 issues in 2026. CVE-2026-23918, a double-free fixed in 2.4.67, landed in May. In June, the "HTTP/2 Bomb" denial-of-service technique (CVE-2026-49975) hit the default HTTP/2 configurations of several major web servers, Apache included. Apache 2.4.68 and 2.4.69 fixed 13 and 20 CVEs respectively. HTTP/2 brings real performance benefits, but its concurrency makes the code that handles it complex, and complex concurrent code is where bugs like this one live.
The way this flaw was found is also worth noting. It was reported on October 26 by Anthropic's OSS Scanner, an AI-driven tool for finding flaws in open source software, and then reported independently on November 3 by Anthony Weems and Simon Scannell of Google Cloud Vulnerability Research. Two separate teams finding the same subtle race condition within about a week suggests that automated discovery is getting good at exactly the kind of bug that used to take a skilled human a long time to spot. Defenders benefit when those findings go to maintainers first, as they did here. The same capability in other hands means the gap between "a bug exists" and "someone has found it" is shrinking, and so is the time organisations can safely leave a critical patch waiting.
For most organisations, the practical lesson is the same as it was in May: know which of your servers run Apache, know which of them have HTTP/2 enabled, and be able to patch or switch it off quickly. If answering that took more than an hour this time, it is worth fixing before the next one.
Summary
| CVE | CVE-2026-119618 |
| Component | Apache HTTP Server mod_http2, assign_headers() function |
| Type | Race condition leading to heap corruption; potential unauthenticated remote code execution |
| Severity | Critical (Apache); 9.8 (Red Hat); Important for RHEL 9 and 10 packages |
| Affected | Apache HTTP Server 2.4.55 to 2.4.70; standalone mod_h2 2.0.8 to 2.0.42 |
| Condition | mod_http2 loaded and h2 or h2c in Protocols, server-wide or in any virtual host |
| Not affected | HTTP/1.1-only servers; Apache before 2.4.55; RHEL 6, 7, and 8 |
| Fixed | Apache HTTP Server 2.4.71 (November 11, 2026), or vendor backports |
| Workaround | None official; disable HTTP/2 (Protocols http/1.1) or unload mod_http2 |
| Exploitation | None confirmed in the wild; Red Hat cites a reliable exploit in unnamed public research; Wiz reports no exploit available; public PoC causes crashes only |
| Exposure | Wiz: affected versions in 64% of cloud environments, publicly exposed in 51% (configuration not accounted for) |
| Discovered | Anthropic's OSS Scanner (October 26); Anthony Weems and Simon Scannell, Google Cloud Vulnerability Research (November 3) |
| Also fixed in 2.4.71 | CVE-2026-106023 (mod_md, Important); CVE-2026-109163 (mod_vhost_alias, Low) |
| Detection | Worker crashes and restarts; web server process spawning shells or making unexpected outbound connections |
CVE-2026-119618 hits one of the internet's most common web servers, needs no login, and already has public code that can crash it. Whether reliable code execution follows in days or months, the fix is available now, and turning off HTTP/2 is a fast stopgap. Find your Apache servers, check their protocol settings, and patch.
If you need help identifying which of your servers run Apache with HTTP/2 enabled, confirming that vendor backports are actually in place, or monitoring for signs of exploitation, get in touch.