llamaCVEVulnerability Intelligence
/
Advisory · GitHub · Go · github.com/0xJacky/Nginx-UI

GHSA-RXQJ-WW48-X5FV

TestHealthCheck (POST /api/site_navigation/test_health_check/:id) is an auth-only SSRF primitive with a response-body substring oracle, invocable by write-scope service tokens — sibling of fixed GHSA-49r7-63h8-8p86 / GHSA-94h5-jw58-qqq8, not covered by their fixes

Sin parche publicado
Descargar PNG JSON
CVSS 3.1
7.7
Alta
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:N/A:N Fuente: GitHub
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
ParcheSin parcheEl fabricante no ha publicado corrección
Exposición28 dDesde la divulgación, 22 oct 2026, 1:59 a. m.
Detección27 dDesde la divulgación hasta su ingreso al índice
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Explotación
Remediación
Monitoreo
Detectada +1
22 oct 26
18 nov 26
19 nov 26
22 oct 26
29 oct 26
5 nov 26
12 nov 26
19 nov 26
FechaFaseEventoDetalle
Divulgación Publicada GHSA-RXQJ-WW48-X5FV divulgada · CVSS 7.7 (GitHub)
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés
Summary

The health-check test endpoint accepts a complete attacker-controlled HealthCheckConfig in the request body, replaces the probed URL with the attacker's target_url with no host/IP/port allowlist or redirect restriction, issues the request server-side with attacker-chosen method/headers/body, and returns success = ExpectedStatus-match && ExpectedText-substring-match plus the raw transport error. The route is registered on the plain AuthRequired group (its sibling health-check mutations were moved behind RequireSecureSession; this one was not — the code comment even concedes "this is an arbitrary-fetch primitive"). Result: any authenticated credential, including static write-scope service tokens that cannot complete step-up, gets a server-side arbitrary-fetch primitive usable for internal host/port discovery and, thanks to the body-substring oracle, byte-by-byte extraction of HTTP response content (e.g. cloud instance metadata such as IMDSv1 credentials). Verified on v2.8.4.

### Reporter / Credits:
Sai Sreewathsa Kovalluri, Shon Babu and Kannan K (Researchers), Innspark Solutions https://innspark.in/

Details

Registration: api/sites/router.go:19 —
r.POST("site_navigation/test_health_check/:id", middleware.RejectInDemo(), TestHealthCheck)
on the plain AuthRequired()+Proxy() group. The RequireSecureSession group begins at router.go:21 and DOES contain the sibling mutations (UpdateHealthCheck :31, SyncHealthCheck :32) — the test endpoint is the odd one out. No site_navigation path appears in the service-token restriction lists (internal/middleware/middleware.go), and for write-gating RequireSecureSession is absent entirely here, so a write-scope (or even read+POST path behavior per apiScopeForRequest) service token reaches it.

Handler: api/sites/sitecheck.go:346-391 (TestHealthCheck). It never calls the existing validator validateHealthCheckRequest (that is only used at :198 and :303 for the step-up-gated mutations). The :id path parameter merely selects which stored site's base URL/timeout is used when TargetURL is empty.

Sink: internal/sitecheck/enhanced_checker.go —
- effectiveHealthCheckURL (lines 74-79) replaces the probed URL with config.TargetURL verbatim (no allowlist; rewriteCheckURLScheme only aligns the scheme).
- checkHTTP (line 112+) builds the request with attacker Method/Headers/Body and sends it via client.Do (line 183).
- Response evaluation (lines ~208-236) returns success = statusValid && textValid, plus status_code and the raw transport error string. With attacker-chosen ExpectedStatus (any 100-599), success depends solely on the expected_text / not_expected_text substring match against the response body — a full content oracle, strictly stronger than the blind SSRFs covered by the earlier advisories.

Because the group carries middleware.Proxy(), adding X-Node-ID additionally runs the same fetch from any registered cluster node (see also the middleware-ordering issue reported separately).

Proof of concept
  1. Any valid credential (a write-scope nui_pat_ service token suffices; no 2FA possible/needed).
  2. Have at least one site so a SiteConfig id exists (or enumerate via GET /api/site_navigation).
  3. Substring oracle (verified verbatim on v2.8.4, panel self as an internal target):
  4. Cloud metadata theft: point target_url at http://169.254.169.254/latest/meta-data/
  5. Internal discovery: sweep target_url across RFC1918 hosts/ports and read status_code/error
  6. Cluster variant: add header X-Node-ID: <id> (enumerable via GET /api/nodes) to run the fetch

Suggested fix direction: move the route into the RequireSecureSession group (parity with the GHSA-49r7 fix) and add RequireInteractiveUser mirroring api/external_notify; run validateHealthCheckRequest (extended to reject link-local/loopback/RFC1918 targets and disable redirects) inside TestHealthCheck; stop returning raw transport errors/status codes on this endpoint.

Impact

Server-Side Request Forgery with a response-content oracle. Any holder of an ordinary authenticated credential — including long-lived, static, stealable service tokens intended for automation — can make the server fetch arbitrary internal/external URLs and learn response-body content. Practical consequences: cloud metadata credential theft (IMDSv1/v2), internal service discovery and content reading, and action-relay against internal HTTP services. This is the same class the project rated and fixed for the notification test endpoints (GHSA-49r7-63h8-8p86 / GHSA-94h5-jw58-qqq8); this endpoint is a sibling the fixes never covered, and its content oracle makes it strictly more powerful.

This vulnerability has been manually verified, tested, and confirmed before submission

---

Maintainer consolidation update (2026-10-18)

GHSA-2894-28pm-7h4c, independently submitted by @manus-pi, is consolidated here: it describes the same health-check test route, TargetURL/method/header/body handling, expected-text oracle, and historical missing secure-session gate. The independent contribution is retained; this consolidation does not determine priority beyond the original submission dates.

The missing step-up and original-caller proxy handling were fixed by b6d5d39b2fb00d9e2f51b58d21e3114f84bd80e6, included in v2.9.0. Authorized custom/private-network monitoring and content matching remain intentional capabilities under the management-credential model. Synthetic loopback testing confirms rejection of an MFA-enrolled JWT lacking proof before egress, and permitted management probes with the expected controls. Cloud metadata credential extraction and universal IMDSv2 token recovery were not demonstrated. This is the narrowed accepted scope; the original broader submitted impact language above is not an unconditional maintainer claim.

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:H
AltoBajoNinguno
IntegridadI:N
AltoBajoNinguno
DisponibilidadA:N
AltoBajoNinguno
Versiones

Rangos afectados

Afectada Corregida Última
github.com/0xJacky/Nginx-UI< 2.9.0
v1.9.9
2.9.0
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
gogithub.com/0xJacky/Nginx-UI
Sin parche — — v1.9.9 afectada
Versiones corregidas2.9.0
Explotación

Arsenal público

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

Referencias
Advisoryhttps://github.com/0xJacky/nginx-ui/security/advisories/GHSA-rxqj-ww48-x5fv

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