llamaCVEVulnerability Intelligence
/
Advisory · GitHub · Go · github.com/nezhahq/nezha

GHSA-GRRW-FX36-FV32

Service monitor task type validation allowed scoped Agent task capability escalation

Sin parche publicado
Descargar PNG JSON
CVSS
—
medium
Fuente: —
ExplotaciónSin evidenciaNo figura en CISA KEV
Exploits públicos0Ninguno conocido
ParcheSin parcheEl fabricante no ha publicado corrección
Exposición96 dDesde la divulgación, 15 ago 2026, 12:51 a. m.
Detección96 dDesde la divulgación hasta su ingreso al índice
Timeline

Ciclo de vida de la vulnerabilidad

Hoy
Divulgación
Publicada
Explotación
Remediación
Monitoreo
Detectada +1
15 ago 26
19 nov 26
15 ago 26
8 sept 26
2 oct 26
26 oct 26
19 nov 26
FechaFaseEventoDetalle
Divulgación Publicada GHSA-GRRW-FX36-FV32 divulgada
Monitoreo Detectada Ingresó al índice
Monitoreo Verificada Última verificación contra las fuentes
Resumen
Descripción del advisory
Texto original · GitHub · en inglés

Summary

Nezha Dashboard before 2.3.5 did not restrict a service monitor's task type to the three supported probe types. Service monitors and privileged Agent operations share the same protobuf `Task.Type` namespace, so an authenticated caller with `nezha:service:write` could submit a non-probe type through the service API. The Dashboard could then deliver that task to Agents which the caller was already authorized to manage.

Values corresponding to command execution and Agent configuration could therefore reach those Agent features through the narrower service-monitor API.

Authorization boundary and impact

This issue is not anonymously exploitable and does not bypass tenant or server isolation. Exploitation requires a RoleMember account or PAT with the `nezha:service:write` scope. Existing server ownership checks and PAT `server_ids` restrictions remain enforced, so impact is limited to online Agents already within that user or token's authorization scope; Agents belonging to other users or outside the PAT allowlist cannot be targeted.

Within that existing scope, the issue allowed a service-management credential to invoke Agent capabilities not intended for service monitoring, potentially including command execution as the Agent service account or changing the Agent's Dashboard configuration. The Agent is the authorized execution target and is not itself a separately vulnerable package. These prerequisites and the preserved server authorization boundary make the issue Medium severity.

Affected versions

  • `github.com/nezhahq/nezha` versions before 2.3.5 are affected.
  • Nezha Agent is not independently affected and does not require a separate Agent security release for this issue.

Fix

Dashboard 2.3.5 restricts service monitors to HTTP GET, ICMP ping, and TCP ping task types. The allowlist is enforced at create and update API boundaries, model persistence, scheduler loading, and final task dispatch. Invalid legacy rows are not scheduled or dispatched. Regression tests cover privileged, unknown, and valid probe task types.

Users should upgrade the Dashboard to 2.3.5 or later. Until upgrading, remove `nezha:service:write` from untrusted PATs and RoleMembers.

Fix commit: https://github.com/nezhahq/nezha/commit/38824dbc11a63964c5b9296ae4c68e62b34fa04b

Fixed release: https://github.com/nezhahq/nezha/releases/tag/v2.3.5

Thanks to @Attack-Phantom for reporting the task-type confusion and providing end-to-end reproductions.

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

Versiones

Rangos afectados

Afectada Corregida Última
github.com/nezhahq/nezha< 2.3.5
v1.14.14
2.3.5
Remediación

Estado del parche por paquete

Paquete / productoEstadoCorregida enDesde divulgaciónÚltima versión
gogithub.com/nezhahq/nezha
Sin parche — — v1.14.14 afectada
Versiones corregidas2.3.5
Explotación

Arsenal público

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

Referencias
Advisoryhttps://github.com/nezhahq/nezha/security/advisories/GHSA-grrw-fx36-fv32

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