llamaCVEVulnerability Intelligence
/
Advisory · GitHub · PyPI · frigate

GHSA-W3C4-957Q-2JW4

SSRF via webpush endpoint validation bypass — registration-time check does not match send-time resolution/redirect behavior

Texto original en inglés · Ver en español

Sin parche publicado
Descargar PNG JSON
CVSS 3.1
6.4
Media
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:L/A:N Fuente: GitHub
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
ParcheSin parcheEl fabricante no ha publicado corrección
Exposición70 dDesde la divulgación, 12 sept 2026, 9:13 a. m.
Detección67 dDesde la divulgación hasta su ingreso al índice
Decisión estimadaSeguir
Remediar en la ventana regular de mantenimiento
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 parcial. No hay versión corregida: aplique mitigaciones.
Hoy toca orden, no heroísmo.Esto lo arregla la rutina de siempre.
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Explotación
Remediación
Monitoreo
Detectada +1
12 sept 26
18 nov 26
21 nov 26
12 sept 26
30 sept 26
17 oct 26
4 nov 26
21 nov 26
FechaFaseEventoDetalle
Divulgación Publicada GHSA-W3C4-957Q-2JW4 divulgada · CVSS 6.4 (GitHub)
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés

### Summary
The webpush notification endpoint's SSRF validator
(_validate_push_endpoint, frigate/api/notification.py:31-85, added
in commit ad79e666e) validates the subscription endpoint URL as a
string once at registration time. The actual send path
(frigate/comms/webpush.py:299-322) POSTs to the stored URL verbatim
via pywebpush, which resolves DNS fresh at send time and follows
redirects. Two distinct bypass classes allow any authenticated user,
including restricted/viewer roles, to cause the Frigate server to
send outbound requests to internal or arbitrary hosts.

Details

Bypass class 1 — non-canonical IPv4 forms: the validator uses
ipaddress.ip_address(hostname) to detect private IPs, but glibc's
getaddrinfo (used at actual connection time) accepts forms that
ipaddress.ip_address() rejects as malformed and therefore does not
flag as private.

Live-verified accept/resolve pairs:

Endpoint Validator Resolves to
https://127.0.0.1/fcm/send/tok reject 127.0.0.1
https://127.1/fcm/send/tok ACCEPT 127.0.0.1
https://0x7f.1/fcm/send/tok ACCEPT 127.0.0.1
https://0177.0.0.1/fcm/send/tok ACCEPT 127.0.0.1
https://169.254.169.254.nip.io/fcm/send/tok ACCEPT 169.254.169.254

Bypass class 2 — redirect escape: an attacker registers their own
legitimately valid public HTTPS endpoint (passes all registration-time
checks). When Frigate later sends a notification, the attacker's
server responds with a 302 redirect. The actual send call follows the
redirect to any scheme/host/port — the HTTPS/port-443 constraints only
exist in the registration-time validator, not the send-time request.
This converts a blind SSRF into attacker-steered GET/POST requests
against arbitrary internal URLs.

Test coverage gap: test_webpush_registration.py tests only the pure
validator function in isolation — it correctly pins rejection of
canonical IP literals, .local/.internal/.home.arpa suffixes, wrong
ports, and embedded credentials. It contains zero tests for
non-canonical IPv4 encodings, zero for hostname-to-IP resolution
behavior, and zero for redirect behavior at send time. Every rejected
case in the test suite has an accepted sibling that behaves identically
on the wire.

PoC

Live-tested with a real frigate.api.notification.router mounted in
FastAPI, a real User model on SQLite, real HTTP via TestClient, as an
authenticated user with restricted role "patrol":

Step 1 — control, canonical private IP correctly rejected:
POST /notifications/register
{"sub": {"endpoint": "https://127.0.0.1/fcm/send/tok", ...}}
-> 400 {"success": false, "message": "Invalid subscription: endpoint
must not use a private address"}

Step 2 — bypass, non-canonical form accepted:
POST /notifications/register
remote-user: restricted, remote-role: patrol
{"sub": {"endpoint": "https://127.1/fcm/send/tok", ...}}
-> 200 {"success": true, "message": "Successfully saved token."}

Step 3 — persisted:
[{"endpoint": "https://127.1/fcm/send/tok", ...}]

Step 4 — send path attempted with valid EC keys (real pywebpush call):
ConnectionError: HTTPSConnectionPool(host='127.1', port=443)...
[Errno 111] Connection refused

The connection-refused error confirms the send attempt actually
dialed the loopback-resolved address 127.0.0.1 — the SSRF payload was
delivered to the intended internal target, with only ordinary user
registration privilege required.

### Impact
Any authenticated user, including restricted/viewer-role accounts,
can register a webpush subscription that causes the Frigate server to
make outbound requests to internal network addresses or, via the
redirect bypass, to attacker-steered arbitrary destinations. The
outbound requests carry a JSON body containing camera names,
detected-object labels, zones, GenAI summaries, event IDs, and
thumbnail paths. Because the attacker supplies their own p256dh/auth
encryption keys during registration, they can decrypt any response
data captured. Internal state-changing GET endpoints reachable via
the redirect bypass are also exposed.

### Suggested Fix
Resolve the hostname at validation time via socket.getaddrinfo() and
apply ipaddress.is_global() to every resolved address rather than
parsing the literal hostname string. Additionally, either re-validate
the resolved address at send time, or construct the pywebpush request
using a session with redirects disabled
(requests.Session(); session.post(..., allow_redirects=False)).
Consider purging or flagging pre-existing stored subscriptions that
fail the new check on upgrade.

Please credit this report to Mayssare Amakhtari (GitHub: @cy3erm).

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:C
CambiadoSin cambio
ConfidencialidadC:L
AltoBajoNinguno
IntegridadI:L
AltoBajoNinguno
DisponibilidadA:N
AltoBajoNinguno
Score de explotaciónQué tan fácil y probable es que la exploten
8.7/ 10Muy alto
Vector 8.7 (muy fácil) · sin evidencia de explotación.
Versiones

Rangos afectados

Afectada Corregida Última
frigate≥ 0.18.0-rc1
0.7.0
0.18.0-rc1
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
pypifrigate
Sin parche — — 0.7.0 no afectada
Explotación

Arsenal público

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

Referencias
Advisoryhttps://github.com/blakeblackshear/frigate/security/advisories/GHSA-w3c4-957q-2jw4

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