Summary
YesWiki Vulnerable to Authenticated PHP Object Injection in BazarImportAction via unserialize
Full technical description
Details
Sink
tools/bazar/services/CSVManager.php line 372-399:
public function importEntry(array $importedEntries, string $formId): ?array
{
if (!$this->importdone) {
// ...
foreach ($importedEntries as $entry) {
$entry = unserialize(base64_decode($entry)); // <-- SINK
$entry = array_map('strval', $entry);
// ...
There is no ['allowed_classes' => false] argument; arbitrary classes are instantiated. The subsequent array_map('strval', $entry) additionally exercises __toString on each top-level array element, doubling the magic-method surface available to a gadget chain.
Source
tools/bazar/actions/BazarImportAction.php:
// formatArguments()
'mode' => (isset($_POST['submit_file']) && !empty($_FILES['fileimport']['name'])) ? 'submitfile' :
(isset($_POST['importfiche']) ? 'importentries' : 'default'),
'importentries' => $_POST['importfiche'] ?? null,
// run()
case 'importentries':
// ...
$importedEntries = $this->CSVManager->importEntry($this->arguments['importentries'], $vID['id']);
break;
$_POST['importfiche'] flows directly to the sink. The mode switches to 'importentries' whenever the request body contains the key, so an attacker need only POST importfiche[0]=<payload>.
Reachability
The action is registered as
bazarimport. The defaultBazaRpage (setup/sql/default-content.sql->BazaRpage entry, ships with{{bazar showexportbuttons="1"}}) routes?BazaR&vue=importer&id_typeannonce=<N>toBazarAction::run()->case VOIR_IMPORTER -> callAction('bazarimport', ...)(tools/bazar/actions/BazarAction.php:257-258). So the sink is reachable on a default install with no extra page authoring.BazarImportAction::run()calls$this->checkSecuredACL()with the default$adminOnly=true. Only wiki admins (or accounts the admin has added to thebazarimportaction ACL) can execute it.The
importentriesbranch does NOT invokeCsrfTokenController::checkToken(...). Greppingtools/bazar/actions/BazarImportAction.phpconfirms the action class has nocsrforcheckTokenreference at all. This is asymmetric with sibling actions:tools/bazar/controllers/FormController.phpdoes callcheckToken('main', 'POST', 'confirmDeleteToken')for destructive operations. The import path skips the same protection.Therefore the full kill chain for a remote attacker is:
a. Identify any admin user on the target wiki.
b. Deliver an HTML page (email, chat, link) that auto-POSTsimportfiche[0]=<base64-encoded PHPGGC payload>tohttps://<wiki>/?BazaR&vue=importer&id_typeannonce=1.
c. The admin's session cookie is sent automatically; the action passescheckSecuredACL; the unserialize fires.
Gadget chain availability
composer.json requires doctrine/annotations ^1.11 and doctrine/cache ^1.10. Both have published PHPGGC chains (Doctrine/RCE1, Doctrine/FW1, Doctrine/FW2, etc., from https://github.com/ambionics/phpggc). These chains terminate in either system($cmd) (RCE1) or file_put_contents($php_file, $contents) (FW1) entry-points -- both sufficient to give the attacker shell on the YesWiki host.
This advisory does not include a working PHPGGC chain end-to-end (writing a chain that survives YesWiki's exact dependency-resolved class graph is separate work). The PoC demonstrates the primitive (attacker-controlled class instantiation + magic-method execution); the chain is a downstream exercise using public tooling.
Past advisories cross-check
YesWiki's published GitHub advisories cover XSS, SQLi, arbitrary-PHP-file-write RCE, path traversal, and unauthenticated backup download. None covers an unserialize / PHP-object-injection sink, so this is a novel vulnerability class for the project.
PoC
A self-contained PoC reproducing the inner loop is available; it copies the exact two-line sink and proves that attacker-controlled __destruct runs without booting the full application.
Run:
php poc.php
Output (verbatim):
Crafted importfiche[0] payload (form-ready, urlencoded):
YToxOntpOjA7Tzo2OiJHYWRnZXQiOjE6e3M6NjoibWFya2VyIjtzOjIyOiJQV05FRC1GUk9NLVVOU0VSSUFMSVpFIjt9fQ%3D%3D
== before importEntry ==
[Gadget] __destruct fired with marker='PWNED-FROM-UNSERIALIZE'
PHP Fatal error: Uncaught Error: Object of class Gadget could not be converted to string ...
[Gadget] __destruct fired with marker='PWNED-FROM-UNSERIALIZE'
The two [Gadget] __destruct fired lines (one from inside the loop, one from the engine shutdown after the TypeError) confirm that the attacker-defined Gadget::__destruct executed -- with the attacker-supplied marker -- inside the unmodified importEntry code path.
End-to-end against a live YesWiki install:
curl -i -b "yeswiki_session=<admin_cookie>" \
-X POST "https://wiki.example.com/?BazaR&vue=importer&id_typeannonce=1" \
--data-urlencode \
"importfiche[0]=YToxOntpOjA7Tzo2OiJHYWRnZXQiOjE6e3M6NjoibWFya2VyIjtzOjIyOiJQV05FRC1GUk9NLVVOU0VSSUFMSVpFIjt9fQ=="
(replace the payload with a real PHPGGC Doctrine/FW1 or Doctrine/RCE1 output to obtain RCE on the target host).
Impact
- Authenticated wiki admin who lands on attacker-controlled HTML obtains remote code execution on the YesWiki server (via the cross-site forgery path; no admin interaction with the import UI is required).
- An attacker who has already compromised an admin password upgrades from "wiki content management" to "OS shell on the hosting box".
- The compromise survives the wiki layer entirely: the attacker can write web shells, exfiltrate other sites on shared hosting, modify
wakka.config.php, dump the MySQL database, and pivot from there.
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.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
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. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
tools/bazar/services/CSVManager.php::importEntry-- pass['allowed_classes' => false]tounserialize, or, better, replace the base64+serialize transport with the JSON transport the current UI already uses (?api/entries/{formId}POST intools/bazar/presentation/javascripts/bazar-import.js). The serialized-PHP transport appears to be an unused legacy path.tools/bazar/actions/BazarImportAction.php-- add aCsrfTokenController::checkToken('main', 'POST', 'csrf-token', false)guard for the'importentries'mode (and any other state-changing modes). The existingtools/bazar/controllers/FormController.phppattern can be lifted directly.
Frequently Asked Questions
- What is CVE-2026-52777? CVE-2026-52777 is a critical-severity cross-site request forgery (CSRF) vulnerability in yeswiki/yeswiki (composer), affecting versions < 4.6.6. It is fixed in 4.6.6. A victim's authenticated browser session is used to submit forged requests to an application that cannot distinguish them from legitimate ones.
- Which versions of yeswiki/yeswiki are affected by CVE-2026-52777? yeswiki/yeswiki (composer) versions < 4.6.6 is affected.
- Is there a fix for CVE-2026-52777? Yes. CVE-2026-52777 is fixed in 4.6.6. Upgrade to this version or later.
- Is CVE-2026-52777 exploitable, and should I be worried? Whether CVE-2026-52777 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
- What actually determines whether CVE-2026-52777 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.
- How do I fix CVE-2026-52777? Upgrade
yeswiki/yeswikito 4.6.6 or later.