Auditing Google Cloud Platform IAM permissions to identify overly permissive bindings, primitive role usage, service account key proliferation, and cross-project access risks using gcloud CLI, Policy Analyzer, and IAM Recommender.
Instrucciones de origen · Vista previa de solo lectura
name
auditing-gcp-iam-permissions
description
Auditing Google Cloud Platform IAM permissions to identify overly permissive bindings, primitive role usage, service account key proliferation, and cross-project access risks using gcloud CLI, Policy Analyzer, and IAM Recommender.
When performing security assessments of GCP organization or project IAM configurations
When identifying service accounts with excessive permissions or unused access
When compliance requirements mandate review of access controls and role assignments
When investigating potential lateral movement through IAM misconfigurations
When reducing the blast radius of compromised credentials by scoping down permissions
Do not use for VPC firewall rule auditing (use network security tools), for GKE RBAC auditing (use Kubernetes-specific RBAC tools), or for real-time threat detection on IAM actions (use SCC Event Threat Detection).
Detection Gaps & Validation
search-all-iam-policies only sees allow bindings: it does not return IAM Conditions, deny policies, or the effective result of inheritance. A binding listed at the project may be overridden (or a privilege granted) by an org/folder-level binding you didn't enumerate. Always resolve effective access with gcloud asset analyze-iam-policy.
Impersonation = invisible privesc:roles/iam.serviceAccountTokenCreator, roles/iam.serviceAccountUser, and iam.serviceAccounts.getAccessToken/actAs let a low-priv principal act as a high-priv SA. Chains (A→B→C) won't show in any single binding — walk the graph.
Domain-wide delegation is set in the Workspace Admin console, not GCP IAM — gcloud cannot see it, so a DWD-enabled SA looks benign. Cross-check the Admin SDK / clientId OAuth grants separately.
Recommender blind spots: IAM Recommender needs ~90 days of usage data; new or rarely-used SAs produce no recommendation (not "least privilege confirmed"). Basic roles (Owner/Editor/Viewer) hide hundreds of service permissions a wildcard grep won't expand.
Validate effective access: confirm a who-can claim with gcloud asset analyze-iam-policy --identity=... --full-resource-name=... using --expand-groups/--analyze-service-account-impersonation; before removing a binding, verify in Cloud Audit Logs that the principal hasn't used it recently.
Prerequisites
GCP organization or project with roles/iam.securityReviewer and roles/cloudAsset.viewer
gcloud CLI authenticated with appropriate permissions
Cloud Asset API enabled (gcloud services enable cloudasset.googleapis.com)
IAM Recommender API enabled (gcloud services enable recommender.googleapis.com)
Policy Analyzer API enabled (gcloud services enable policyanalyzer.googleapis.com)
Workflow
Step 1: Enumerate IAM Bindings Across the Organization
List all IAM bindings at organization, folder, and project levels to understand the full access landscape.
# Organization-level IAM bindings
gcloud organizations get-iam-policy ORG_ID \
--format=json > org-iam-policy.json
# Search all IAM policies across the organization
gcloud asset search-all-iam-policies \
--scope=organizations/ORG_ID \
--format="table(resource, policy.bindings.role, policy.bindings.members)" \
--limit=500
# Find all users and service accounts with Owner role
gcloud asset search-all-iam-policies \
--scope=organizations/ORG_ID \
--query="policy:roles/owner" \
--format="table(resource, policy.bindings.members)"# Find all bindings using primitive roles (Owner, Editor, Viewer)
gcloud asset search-all-iam-policies \
--scope=organizations/ORG_ID \
--query="policy:roles/owner OR policy:roles/editor" \
--format=json | python3 -c "
import json, sys
data = json.load(sys.stdin)
for result in data:
resource = result.get('resource', '')
for binding in result.get('policy', {}).get('bindings', []):
role = binding.get('role', '')
if role in ['roles/owner', 'roles/editor']:
for member in binding.get('members', []):
print(f'{resource} | {role} | {member}')
"
Step 2: Audit Service Accounts and Their Keys
Identify service accounts with excessive permissions, user-managed keys, and unused accounts.
# List all service accounts in a project
gcloud iam service-accounts list \
--project=PROJECT_ID \
--format="table(email, displayName, disabled)"# Check for user-managed keys (should be minimized)for sa in $(gcloud iam service-accounts list --project=PROJECT_ID --format="value(email)"); do
keys=$(gcloud iam service-accounts keys list \
--iam-account="$sa" \
--managed-by=user \
--format="table(name.basename(),validAfterTime,validBeforeTime)")
if [ -n "$keys" ]; thenecho"=== $sa ==="echo"$keys"fidone# Find service accounts with admin roles across all projects
gcloud asset search-all-iam-policies \
--scope=organizations/ORG_ID \
--query="policy.bindings.members:serviceAccount AND (policy:roles/owner OR policy:roles/editor OR policy:admin)" \
--format="table(resource, policy.bindings.role, policy.bindings.members)"# Check service account IAM policies (who can impersonate)for sa in $(gcloud iam service-accounts list --project=PROJECT_ID --format="value(email)"); doecho"=== $sa ==="
gcloud iam service-accounts get-iam-policy "$sa" --format=json 2>/dev/null
done
Step 3: Use IAM Recommender to Identify Excess Permissions
Leverage GCP's IAM Recommender to find roles that grant more access than actually used.
# List IAM role recommendations for a project
gcloud recommender recommendations list \
--project=PROJECT_ID \
--recommender=google.iam.policy.Recommender \
--location=global \
--format="table(name, description, priority, stateInfo.state)"# Get detailed recommendation
gcloud recommender recommendations describe RECOMMENDATION_ID \
--project=PROJECT_ID \
--recommender=google.iam.policy.Recommender \
--location=global \
--format=json
# List insights about IAM usage
gcloud recommender insights list \
--project=PROJECT_ID \
--insight-type=google.iam.policy.Insight \
--location=global \
--format="table(name, description, severity, category)"# Apply a recommendation (after review)
gcloud recommender recommendations mark-claimed RECOMMENDATION_ID \
--project=PROJECT_ID \
--recommender=google.iam.policy.Recommender \
--location=global \
--etag=ETAG
Step 4: Analyze Effective Permissions with Policy Analyzer
Use Policy Analyzer to determine effective access for specific principals or resources.
# Check who has access to a specific resource
gcloud asset analyze-iam-policy \
--organization=ORG_ID \
--full-resource-name="//storage.googleapis.com/projects/_/buckets/sensitive-data-bucket" \
--format="table(identityList.identities, accessControlLists.accesses.role)"# Check what resources a specific user can access
gcloud asset analyze-iam-policy \
--organization=ORG_ID \
--identity="user:developer@company.com" \
--format="table(accessControlLists.resources.fullResourceName, accessControlLists.accesses.role)"# Check who can perform a specific action
gcloud asset analyze-iam-policy \
--organization=ORG_ID \
--full-resource-name="//cloudresourcemanager.googleapis.com/projects/PROJECT_ID" \
--permissions="iam.serviceAccounts.actAs,iam.serviceAccountKeys.create" \
--format="table(identityList.identities, accessControlLists.accesses.permission)"# Find all principals with allUsers or allAuthenticatedUsers access
gcloud asset search-all-iam-policies \
--scope=organizations/ORG_ID \
--query="policy:allUsers OR policy:allAuthenticatedUsers" \
--format="table(resource, policy.bindings.role, policy.bindings.members)"
Step 5: Check for Domain-Wide Delegation and Impersonation Risks
Identify service accounts with domain-wide delegation and impersonation capabilities.
# Check for service accounts with domain-wide delegation# (Requires Admin SDK access to list delegated accounts)
gcloud iam service-accounts list --project=PROJECT_ID --format=json | python3 -c "
import json, sys
accounts = json.load(sys.stdin)
for sa in accounts:
email = sa.get('email', '')
# Check if the SA has domain-wide delegation enabled
# This requires Admin SDK API access
print(f'SA: {email} - Check admin.google.com for delegation status')
"# Find service accounts that other identities can impersonatefor sa in $(gcloud iam service-accounts list --project=PROJECT_ID --format="value(email)"); do
policy=$(gcloud iam service-accounts get-iam-policy "$sa" --format=json 2>/dev/null)
ifecho"$policy" | python3 -c "
import json, sys
p = json.load(sys.stdin)
for b in p.get('bindings', []):
if b['role'] in ['roles/iam.serviceAccountTokenCreator', 'roles/iam.serviceAccountUser']:
print(f' {b[\"role\"]}: {b[\"members\"]}')
" 2>/dev/null; thenecho"=== Impersonation risk: $sa ==="fidone
Step 6: Generate Audit Report and Apply Remediation
Compile findings and implement recommended permission reductions.
# Remove primitive role and replace with predefined role
gcloud projects remove-iam-policy-binding PROJECT_ID \
--member="user:developer@company.com" \
--role="roles/editor"
gcloud projects add-iam-policy-binding PROJECT_ID \
--member="user:developer@company.com" \
--role="roles/compute.viewer"
gcloud projects add-iam-policy-binding PROJECT_ID \
--member="user:developer@company.com" \
--role="roles/storage.objectViewer"# Delete unused service account keys
gcloud iam service-accounts keys delete KEY_ID \
--iam-account=SA_EMAIL
# Disable unused service accounts
gcloud iam service-accounts disable SA_EMAIL --project=PROJECT_ID
Key Concepts
Term
Definition
Primitive Role
Legacy GCP roles (Owner, Editor, Viewer) that grant broad permissions across all services, not recommended for production
Predefined Role
GCP-managed role scoped to specific services and actions, providing more granular access than primitive roles
IAM Recommender
GCP ML-based service that analyzes actual permission usage and suggests role reductions to achieve least privilege
Policy Analyzer
Tool for analyzing effective IAM access across the organization hierarchy, answering who-can-access-what queries
Service Account Key
User-managed credential for service account authentication, a security risk as keys can be exported and do not auto-expire
Domain-Wide Delegation
Grants a service account the ability to impersonate any user in the Google Workspace domain, a significant privilege escalation risk
Tools & Systems
gcloud CLI: Primary tool for querying and managing GCP IAM policies, service accounts, and role bindings
IAM Recommender: ML-based recommendation engine for reducing excessive permissions based on actual usage
Policy Analyzer: Organization-wide effective access analysis tool for understanding who can access what
Cloud Asset Inventory: Cross-project search for IAM policies and resource metadata
ScoutSuite: Multi-cloud auditing tool with GCP IAM-specific checks for role assignments and service accounts
Common Scenarios
Scenario: Reducing Primitive Role Usage Across a GCP Organization
Context: An audit reveals that 60% of IAM bindings across the organization use primitive roles (Owner/Editor). The security team needs to migrate to predefined roles without disrupting developer workflows.
Approach:
Run gcloud asset search-all-iam-policies to inventory all primitive role bindings
Use IAM Recommender to get ML-based suggestions for replacement predefined roles
For each binding, use Policy Analyzer to understand what the principal actually accesses
Create a mapping document: primitive role -> specific predefined roles needed
Apply predefined roles alongside primitive roles for a testing period
Monitor for access denied errors using Cloud Audit Logs
Remove primitive roles after confirming no access issues over 2 weeks
Pitfalls: Primitive roles include permissions across all GCP services, so replacing them requires multiple predefined roles. The Recommender may suggest overly restrictive roles if the observation period does not capture all use cases. Custom roles can fill gaps where no predefined role matches the exact permission set needed.
Output Format
GCP IAM Permissions Audit Report
===================================
Organization: acme-org (ORG_ID: 123456789)
Projects Audited: 25
Audit Date: 2026-02-23
IAM BINDING SUMMARY:
Total bindings: 342
Using primitive roles: 205 (60%)
Using predefined roles: 112 (33%)
Using custom roles: 25 (7%)
CRITICAL FINDINGS:
[IAM-001] Service Account with Owner Role
SA: admin-sa@prod-project.iam.gserviceaccount.com
Role: roles/owner on project prod-project
User-Managed Keys: 3 (oldest: 14 months)
Remediation: Replace with specific predefined roles, delete old keys
[IAM-002] allAuthenticatedUsers Binding
Resource: gs://public-data-bucket
Role: roles/storage.objectViewer
Risk: Any Google account holder can read bucket contents
Remediation: Restrict to specific user groups or service accounts
SERVICE ACCOUNT HEALTH:
Total service accounts: 67
With user-managed keys: 23
Keys older than 90 days: 18
Unused accounts (90+ days): 12
With domain-wide delegation: 2
RECOMMENDER SUGGESTIONS:
Total recommendations: 45
Priority HIGH: 12
Estimated permissions reduced: 2,847 individual permissions
