Skip to content

De confianza en el Reino Unido, EE. UU., la UE e India - respuesta a incidentes 24/7.

GCP Penetration Testing

GCP Penetration Testing

Expose the Google Cloud-specific identity and misconfiguration attack paths before attackers do.

Pruebas manuales por expertosInformes ejecutivosOrientación de remediaciónNueva prueba y atestaciónAnálisis de firmwarePruebas de hardware
GCP Penetration Testing

Visión general

GCP penetration testing assesses Google Cloud Platform environments for the identity, misconfiguration and privilege-escalation vulnerabilities that generic cloud assessments overlook. GCP's IAM model — built around service accounts, workload identity federation, and project/folder/organisation resource hierarchy — creates attack surfaces distinct from AWS and Azure, and testing must be tailored to it. Common GCP attack vectors include service account key exposure: GCP allows the creation of long-lived JSON key files for service accounts, and these keys — when accidentally committed to source code, baked into container images or left in Cloud Storage — provide persistent access that bypasses all other IAM controls. IntelligenceX scans for exposed keys across public Cloud Storage buckets, container registries, Cloud Source Repositories and common secret-leakage paths. IAM over-permission is the second major vector. GCP's primitive roles — Viewer, Editor and Owner — are frequently assigned at the project level, granting broad permissions across all services. We identify all project-level primitive role bindings, all service accounts with Owner or Editor, all cross-project service account impersonation chains, and all workload identity pool configurations that could allow an external identity to impersonate a privileged service account. The GCE metadata endpoint (http://metadata.google.internal/computeMetadata/v1/) is the most commonly exploited GCP entry point. Any Compute Engine instance reachable from an application with an SSRF vulnerability can be used to retrieve the instance's access token, which grants whatever permissions the instance's attached service account holds. We test all SSRF vectors against the metadata endpoint and document the full privilege chain available via the instance token. GCP-specific testing also covers: Cloud Storage bucket IAM bindings for public access (allUsers or allAuthenticatedUsers); Cloud Functions and Cloud Run service URLs for unauthenticated invocation; Cloud SQL instances with public IP and weak password policies; Google Kubernetes Engine cluster access controls and workload identity binding; BigQuery dataset access for public read; Cloud KMS key IAM policies; and organisation policy constraints that should prevent public resource exposure but are misconfigured. All findings are mapped to the CIS Google Cloud Platform Foundation Benchmark and MITRE ATT&CK for Cloud (the GCP-specific techniques under Persistence, Privilege Escalation, Credential Access and Collection). The report includes exploited attack chains with step-by-step proof of concept, GCP-native remediation guidance and an attestation letter for auditors.

Metodología y estándares

CIS Google Cloud Platform Foundation Benchmark, MITRE ATT&CK for Cloud (GCP techniques), PTES and NIST SP 800-115. Google Cloud testing policies confirmed during scoping.

Qué incluye

Service account key exposure scan: source code, container images, Cloud Storage, Cloud Source Repositories
IAM primitive role (Owner/Editor/Viewer) audit at project, folder and organisation level
GCE metadata endpoint SSRF exploitation for instance credential theft
Workload identity federation misconfiguration and cross-project impersonation chain analysis
Cloud Storage bucket IAM audit: allUsers/allAuthenticatedUsers bindings and public data exposure
Cloud Functions and Cloud Run unauthenticated invocation testing
GKE cluster access control, workload identity and RBAC review
Cloud KMS key policy and Cloud Secret Manager access review
BigQuery dataset public access and data-exfiltration path analysis
Organisation policy constraint gap analysis and project-level inheritance review

Qué recibe

Exploited GCP attack chains with step-by-step proof of concept
CIS GCP Benchmark and MITRE ATT&CK for Cloud mapped findings with CVSS v4.0 severity
IAM over-permission and service account key exposure findings with reproduction steps
GCP-native remediation guidance (IAM conditions, VPC Service Controls, org policy constraints)
Executive risk summary and prioritised remediation roadmap
Retest of remediated findings and attestation letter
Alineado con OWASPInformes ejecutivosOrientación de remediaciónNueva prueba incluidaCarta de atestaciónSin volcados de escáner

Preguntas frecuentes

GCP's IAM model is fundamentally different from AWS and Azure: service accounts are both identities and resources, workload identity federation replaces service account keys in modern architectures but is itself misconfigurable, and GCP's resource hierarchy (organisation > folder > project) creates inheritance paths that can silently grant broad permissions from parent levels. A generic cloud assessment misses these GCP-specific attack surfaces. We test the GCP identity model in depth, not just standard cloud misconfigurations.

A GCP service account key is a long-lived JSON credential file that authenticates as a service account from anywhere on the internet. Unlike AWS temporary credentials, GCP service account keys do not expire automatically — a key created for a one-off task can remain valid for years. Keys exposed via source code, container images or Cloud Storage provide persistent, remote access that is easy to miss in standard access reviews. We scan all likely exposure paths and test whether exposed keys still carry exploitable permissions.

We review the GKE cluster configuration for public endpoint exposure, authorised networks, legacy ABAC (Attribute-Based Access Control), workload identity binding (whether pods use the Compute Engine default service account versus namespace-specific Kubernetes service accounts), network policies, pod security standards, and RBAC over-assignment. Where the cluster is accessible, we attempt privilege escalation from a compromised pod to the node's service account token and then to GCP project-level permissions.

Yes. Our report maps findings to control requirements (SOC 2 CC6.1/CC6.6/CC6.8 for infrastructure security; ISO 27001 Annex A 8.8 for vulnerability management and 8.29 for security testing in development) and the attestation letter confirms scope and remediation status for auditor review.

Yes, and this is the most realistic engagement. An application-and-GCP combined test maps how an SSRF vulnerability in a Cloud Run service or App Engine application reaches the metadata endpoint, retrieves the service account token, and escalates to project-level access or lateral movement into Cloud Storage or BigQuery. This mirrors the actual attack chains we see in GCP incident reports.

Hable hoy mismo con un experto en seguridad

Una prueba de penetración, una auditoría o monitorización 24/7: nuestro equipo está preparado en el Reino Unido, EE. UU., la UE e India.