18,500+

requests at staged installer paths in the first 21 hours, per Wordfence

12,632

exploit attempts Wordfence blocked in one 24 hour period

25

release branches WordPress patched, back to 4.7

6 Nov

federal deadline CISA set to apply the fix

The update is the attack surface

The most dangerous moment in the life of a WordPress site is now, briefly, the moment it patches itself. CVE-2026-119977 turns the core updater into an entry point: while a full upgrade runs, a fresh copy of WordPress sits unpacked in a folder the web server will serve to anyone, and the first visitor to reach it can take control of the server.

Every security program tells its owners to update quickly. That advice is still right, and it is the fix here too. But this bug shows that the update process is code like any other, with its own inputs, files and timing, and it runs on every WordPress site, mostly with nobody watching. Most organizations test the WordPress they run. Very few test what WordPress does while it replaces itself.

The vendor rated the flaw Low. CISA lists it as exploited, with a federal deadline of 6 Nov 2026. Both judgements are defensible, and the gap between them is the lesson. A race that is hard to win on one site is easy to attempt on every site at once, and attackers do not need to win often.

We have written before about the trust an update channel carries. In April, a WordPress plugin with 800,000 installs had its update system hijacked and pushed a remote access toolkit to every site that clicked update. That was an attack on the content of an update. This one is an attack on the process, and the code that ships is genuine.

What is WordPress CVE-2026-119977?

CVE-2026-119977 is a time of check, time of use race condition in WordPress core, tracked by CISA as CWE-367. The WordPress security advisory, GHSA-4w3m-c66w-69mg, describes it plainly: during a core update that uses a complete package, WordPress extracted the new files to a predictable directory inside the web served upgrade folder, and an unauthenticated attacker who wins a race against that extraction could reach remote code execution.

No account is needed. CISA's alert says the flaw could allow unauthenticated attackers to execute code on the server during the upgrade window, and attributes it to predictable, web accessible staging paths used during full package core upgrades. The window is the only precondition, and every site that updates opens one.

The timeline is short. WordPress released 7.1.4 on 28 Oct 2026 with the fix and published the advisory on 29 Oct. The advisory says the fix was backported to all branches back to 4.7, which is 25 release branches in all, from 4.7.39 up to 7.1.4. CISA added the CVE to its Known Exploited Vulnerabilities catalog on 3 Nov and gave federal civilian agencies until 6 Nov to apply it. KEV marks known ransomware use as unknown.

CVE-2026-119977
  • Product //

    WordPress core, the updater
  • Affected //

    4.7.0 to 7.1.3, across 25 branches
  • Fixed in //

    7.1.4, 7.0.8, 6.9.11 and the matching release on each older branch
  • Authentication //

    none
  • Vendor severity //

    Low
  • Exploited //

    Yes, per CISA KEV, added 3 Nov 2026

The exploitation numbers come from Wordfence, as reported by The Hacker News and Qualys. In the first 21 hours after public disclosure, Wordfence recorded more than 18,500 requests aimed at staged setup-config.php installer paths. Later it reported blocking 12,632 exploit attempts in a single 24 hour period. Those are attempts against sites behind one firewall product. They are not confirmed compromises, and they are not a count of the whole internet. They do show that scanning started almost at once and kept going.

The Hacker News also reports a CVSS score of 8.1 for the flaw without naming who scored it, while the WordPress advisory itself rates severity Low. We use the vendor's rating here and explain the difference in section four.

Sites older than 4.7 sit outside the fix. The WordPress release notes say 4.6 and earlier no longer receive security updates. Those sites are rare, and they have bigger problems, but anyone who still runs one should treat this as one more reason to move.

How a core upgrade becomes code execution

The mechanism matters more than the exploit, so we describe it at the level of what each stage achieves.

WordPress unpacks update packages under wp-content/upgrade. The developer reference for the updater shows the working folder is built inside that directory, and the advisory confirms that the folder is web served and the path predictable. For the length of the extraction and copy, a complete WordPress release sits inside the site's document root.

A freshly unpacked copy of WordPress has no configuration file. WordPress handles that case with its installer: the official install guide says that when WordPress cannot find wp-config.php, it offers to create one, and that loading wp-admin/setup-config.php in a browser starts that process. On a new site that is the friendly first run screen. Inside a staging folder on a live server, it is an installer that answers whoever arrives first.

That is why the requests Wordfence counted were aimed at setup-config.php inside staged paths. An attacker who reaches the staged installer before the updater finishes can configure that copy of WordPress on their own terms, and from an administrator session on any WordPress install, running code on the server is a feature, not a bug. The server is the same one the real site runs on.

Figure 01

From an update to a shell

  1. 01

    Wait //

    The attacker polls predictable staging paths across many sites.

  2. 02

    Window //

    A site starts a full core upgrade and unpacks it in a web folder.

  3. 03

    Race //

    The attacker reaches the staged installer first.

  4. 04

    Execute //

    Control of the staged copy becomes code on the server.

Result //

The site patched itself and was compromised doing it.

No credential and no plugin bug are involved. The only input is timing.

Two details shape the risk. First, the attacker does not need to know when a site will update. Polling is cheap, and the request volume Wordfence reported is what polling looks like. Second, the upgrade that opens the window is the one administrators are told to run. The WordPress release post notes that sites with automatic background updates begin the update process on their own. That is good for patching speed, and it also means many sites run an upgrade in the same few days after every release, which is when attackers know to look.

The advisory scopes the flaw to updates that use a complete package. WordPress does not document in this advisory which updates use the complete package and which use smaller partial packages, so we do not assume that any given minor update is safe from it. The safe assumption is that a vulnerable updater may open the window, and the fix is to stop running a vulnerable updater.

Why a Low rating and a KEV listing can both be right

On one site, the window is short. Qualys describes exploitation as possible only during a short window while a complete core update is in progress, and a site runs that update a few times a year. A rating built on the likelihood of hitting one site at one moment comes out low, and the WordPress advisory says Low.

Across the web, the window is constant. Every WordPress site on the internet opens a window each time it upgrades. An attacker who sends a few requests to every site they can find, every few minutes, does not care which site is in its window. Some site always is. The 18,500 requests in 21 hours are the cost of that strategy, and it is cheap.

That is the general lesson for anyone who triages vulnerabilities by score. Severity ratings describe a single instance. Attackers work at the level of the population. A race condition with a narrow window, an exposure that only exists during maintenance, a default that only matters on a fresh install: each looks minor on one host and becomes a steady source of compromises across a million hosts. We see the same pattern on engagements, where the finding that lands us inside is rarely the highest scored item in the last scan report.

None of this means the vendor got it wrong. A Low rating told administrators the truth about any one site. CISA's listing told them the truth about the internet. If your process only reads one of those signals, it will mis-sort the next bug of this shape.

How to detect attacks on the upgrade folder

The upgrade folder has no business serving web requests. The updater writes and reads it through the filesystem, not over HTTP. So any request from the internet for a file under wp-content/upgrade is worth a look, and a request for an installer script there is worth an alert.

Start with your web server or CDN access logs. Search for paths that contain /wp-content/upgrade/. A run of 404 responses shows you were probed, which is expected this week. A 200 response for a PHP file in that folder, and above all for setup-config.php or wp-admin/install.php under it, needs investigation. Line that up against your update history: if the 200 lands during a core upgrade, treat the server as possibly compromised.

Figure 02

Web request for a WordPress installer inside the upgrade folder

Sigma

title: WordPress installer requested inside the core upgrade folder
status: experimental
description: Requests that reach setup-config.php or install.php under wp-content/upgrade, where the core updater stages new releases. Related to CVE-2026-119977.
logsource:
  category: webserver
detection:
  staging:
    cs-uri-stem|contains: '/wp-content/upgrade/'
  installer:
    cs-uri-stem|contains:
      - 'setup-config.php'
      - '/wp-admin/install.php'
  condition: staging and installer
fields:
  - c-ip
  - cs-uri-stem
  - sc-status
falsepositives:
  - None expected; the updater does not use HTTP for this folder
level: high
A starting point to tune against your own telemetry, not a production rule. Drop the installer selection to see all probing of the folder at a lower level.

Logs only help if the request reached them. Sites behind a CDN or hosting firewall may have the probe blocked or logged at the edge only, so check the edge logs as well as the origin.

If you find a successful request, look past the upgrade folder. Check for administrator accounts nobody created, plugins or themes nobody installed, PHP files modified outside a known update, and new scheduled tasks. Our web application persistence audit walks through where attackers leave a way back in on a web server, and the same places apply here. A clean upgrade folder after the fact proves little, because the updater clears that folder before each run.

How we test WordPress update paths on an engagement

On an external test, WordPress sites are often the largest part of the web estate nobody owns. Marketing runs one, a regional office runs another, an agency built a third and still holds the admin password. Our external recon method finds them from DNS, certificate transparency and the address ranges in scope, and our open source tooling on the open source page automates part of that mapping. The first question is not which version each one runs. It is who updates it, how, and whether anyone would notice if an update went wrong.

Then we check the version and patch level of each site we find, and whether the web server will serve files from wp-content/upgrade at all. That check does not require racing anything. A server that refuses web access to the folder is protected against this class of bug whatever the updater does. A server that serves it is relying entirely on the updater being correct.

We do not trigger core upgrades on a client's production site to test the race. That would change the client's system to prove a point, and the finding stands without it: a vulnerable version plus a reachable upgrade folder is the exposure. If a client wants impact demonstrated, we agree a staging copy with them first, in writing.

We also look at the update path as a process. Which sites update automatically, which wait for a person, and which have updates turned off because a plugin broke once in 2023. On many engagements the WordPress sites with updates disabled are older than the ones this CVE affects, and their real exposure is everything published since. Our web application first hour covers how we orient on a site, and our business logic and race checks cover how we think about timing bugs in general: where state is checked in one moment and used in another, and what an outsider can reach in between.

Finally, we check the headers and surface the site presents while we are there. A WordPress install that leaks its version, lists its users through the REST API and serves no security headers tells an attacker it is a quiet target. Our free security headers checker is a quick first look at that last item.

What is still coming

Expect the scanning to continue for months. The probe is a single request to a predictable path, it costs nothing to keep in a botnet's rotation, and the window reopens every time WordPress ships a release. Sites that never apply the fix will keep running a vulnerable updater through every future upgrade.

Expect researchers to look harder at other updaters. The pattern here, staging new code inside the document root, is not unique to WordPress. Content management systems, plugin managers and self updating applications all have to put new files somewhere before they swap them in. Where that somewhere is web reachable, the same race exists in principle. We would look at any self updating web application with that question in mind.

Expect a long tail of forgotten installs. The fix reached 25 branches because WordPress chose to backport it as far as 4.7. That helps only sites whose updater still runs. A site with updates switched off, or pinned to a host's old image, is now on a public exploited list with no automatic route to the fix.

The durable control is not a version number. It is a web server rule that the upgrade folder is never served, applied to every WordPress site you run, including the ones nobody remembers owning.

What to do today

  1. Confirm every WordPress site runs 7.1.4 or the patched release for its branch. The fixed releases run from 4.7.39 to 7.1.4; anything older on a branch is vulnerable, and 4.6 or earlier gets no fix.

  2. Block web access to wp-content/upgrade at the web server. The updater reads and writes that folder through the filesystem, so a deny rule there costs nothing and closes this class of bug.

  3. Find the WordPress sites nobody owns. Agency builds, regional sites and old campaign pages are where updates stopped, and where this CVE will live longest.

  4. Search access logs for requests into wp-content/upgrade. Treat a 200 response for an installer script there as a possible compromise. The Sigma rule above is a starting point.

  5. If a request succeeded, audit for persistence. Look for new administrators, unknown plugins, modified PHP files and new scheduled tasks, not only the upgrade folder.

  6. Read exploited lists alongside vendor severity. A Low that appears on KEV should move to the front of the queue.