Skip to main content
Pulumi logo Pulumi logo
  1. Docs
  2. Reference
  3. Pre-built Policy Packs
  4. CIS
  5. Google Cloud

CIS 8.1 - Google Cloud

    This page lists all 102 policies in the CIS 8.1 pack for Google Cloud, as published in cis-google-cloud version 1.0.3.

    Policies by control

    1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.

    1.2 — Ensure that a process exists to address unauthorized assets on a weekly basis. The enterprise may choose to remove the asset from the network, deny the asset from connecting remotely to the network, or quarantine the asset.

    2.2 — Ensure that only currently supported software is designated as authorized in the software inventory for enterprise assets. If software is unsupported, yet necessary for the fulfillment of the enterprise’s mission, document an exception detailing mitigating controls and residual risk acceptance. For any unsupported software without an exception documentation, designate as unauthorized. Review the software list to verify software support at least monthly, or more frequently.

    3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.

    3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.

    3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).

    3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.

    4.1 — Establish and maintain a documented secure configuration process for enterprise assets (end-user devices, including portable and mobile, non-computing/IoT devices, and servers) and software (operating systems and applications). Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.

    4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.

    5.4 — Restrict administrator privileges to dedicated administrator accounts on enterprise assets. Conduct general computing activities, such as internet browsing, email, and productivity suite use, from the user’s primary, non-privileged account.

    7.1 — Establish and maintain a documented vulnerability management process for enterprise assets. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.

    7.3 — Perform operating system updates on enterprise assets through automated patch management on a monthly, or more frequent, basis.

    8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.

    8.3 — Ensure that logging destinations maintain adequate storage to comply with the enterprise’s audit log management process.

    8.5 — Configure detailed audit logging for enterprise assets containing sensitive data. Include event source, date, username, timestamp, source addresses, destination addresses, and other useful elements that could assist in a forensic investigation.

    11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.

    11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.

    12.1 — Ensure network infrastructure is kept up-to-date. Example implementations include running the latest stable release of software and/or using currently supported network as a service (NaaS) offerings. Review software versions monthly, or more frequently, to verify software support.

    12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.

    13.1 — Centralize security event alerting across enterprise assets for log correlation and analysis. Best practice implementation requires the use of a SIEM, which includes vendor-defined event correlation alerts. A log analytics platform configured with security-relevant correlation alerts also satisfies this Safeguard.

    16.1 — Establish and maintain a secure application development process. In the process, address such items as: secure application design standards, secure coding practices, developer training, vulnerability management, security of third-party code, and application security testing procedures. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.

    16.12 — Apply static and dynamic analysis tools within the application life cycle to verify that secure coding practices are being followed.

    Policy details

    ai-platform-endpoint-configuration-kms-key-configured

    Severity: medium · Enforcement: advisory

    Ensure AI Platform endpoint configuration uses Cloud KMS

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure the encryptionSpec property with a kmsKeyName when creating AI Platform endpoints. Ensure the KMS key is in the same region as the endpoint and grant the Vertex AI service account access to the key.

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    // Create a KMS key for endpoint encryption
    const keyRing = new gcp.kms.KeyRing("ai-keyring", {
        location: "us-central1",
    });
    
    const cryptoKey = new gcp.kms.CryptoKey("ai-endpoint-key", {
        keyRing: keyRing.id,
        rotationPeriod: "7776000s", // 90 days
    });
    
    // Create an AI Platform endpoint with CMEK
    const endpoint = new gcp.vertex.AiEndpoint("my-endpoint", {
        location: "us-central1",
        displayName: "My Secure Endpoint",
        // Configure customer-managed encryption key
        encryptionSpec: {
            kmsKeyName: cryptoKey.id,
        },
    });
    

    ai-platform-notebook-instance-kms-key-configured

    Severity: medium · Enforcement: advisory

    Ensure AI Platform notebook instance uses Cloud KMS

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure the kmsKey property for notebook instances or virtualMachine.virtualMachineConfig.encryptionConfig.kmsKey for runtimes. Grant the Notebooks service account encrypt/decrypt permissions on the KMS key.

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    // Create a KMS key for notebook encryption
    const keyRing = new gcp.kms.KeyRing("notebook-keyring", {
        location: "us-central1",
    });
    
    const cryptoKey = new gcp.kms.CryptoKey("notebook-key", {
        keyRing: keyRing.id,
        rotationPeriod: "7776000s", // 90 days
    });
    
    // Option 1: Notebook Instance with CMEK
    const notebookInstance = new gcp.notebooks.Instance("my-notebook", {
        location: "us-central1-a",
        machineType: "n1-standard-4",
        // Configure customer-managed encryption key
        kmsKey: cryptoKey.id,
        vmImage: {
            project: "deeplearning-platform-release",
            imageFamily: "tf-latest-cpu",
        },
    });
    
    // Option 2: Notebook Runtime with CMEK
    const notebookRuntime = new gcp.notebooks.Runtime("my-runtime", {
        location: "us-central1",
        virtualMachine: {
            virtualMachineConfig: {
                machineType: "n1-standard-4",
                dataDisk: {
                    initializeParams: {
                        diskSizeGb: 100,
                        diskType: "PD_STANDARD",
                    },
                },
                // Configure customer-managed encryption key
                encryptionConfig: {
                    kmsKey: cryptoKey.id,
                },
            },
        },
    });
    

    ai-platform-notebook-no-direct-internet-access

    Severity: high · Enforcement: advisory

    Ensure AI Platform notebook has no direct internet access

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Configure AI Platform notebook instances with noPublicIp: true and deploy within a VPC subnet. Use VPC Service Controls and Private Google Access for controlled connectivity to Google services without public IP addresses.

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    const notebook = new gcp.notebooks.Instance("my-notebook", {
        location: "us-central1-a",
        machineType: "n1-standard-4",
        // Disable public IP to prevent direct internet access
        noPublicIp: true,
        // Deploy in a VPC subnet for network isolation
        network: "projects/my-project/global/networks/my-vpc",
        subnet: "projects/my-project/regions/us-central1/subnetworks/my-subnet",
        // Configure boot disk
        vmImage: {
            project: "deeplearning-platform-release",
            imageFamily: "tf-latest-cpu",
        },
    });
    

    api-gateway-logging-configuration

    Severity: high · Enforcement: advisory

    Ensure API Gateway has proper logging configuration with service accounts and audit logs

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Configure API Gateway with proper logging setup:

    1. Create a dedicated service account for API Gateway backend authentication
    2. Configure ApiConfig with gatewayConfig.backendConfig.googleServiceAccount
    3. Grant the service account appropriate IAM roles for backend access (e.g., roles/run.invoker for Cloud Run)
    4. Enable audit logging for apigateway.googleapis.com service

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    // Step 1: Create service account
    const apiGatewaySA = new gcp.serviceaccount.Account("api-gateway-sa", {
        accountId: "api-gateway-sa",
        displayName: "API Gateway Service Account",
    });
    
    // Step 2: Grant backend access permissions
    const invokerBinding = new gcp.iam.Member("api-gateway-invoker", {
        project: "my-project",
        role: "roles/run.invoker",  // Or roles/cloudfunctions.invoker for Cloud Functions
        member: apiGatewaySA.member,
    });
    
    // Step 3: Configure API Gateway with service account
    const apiConfig = new gcp.apigateway.ApiConfig("my-api-config", {
        api: api.id,
        apiConfigId: "my-config",
        gatewayConfig: {
            backendConfig: {
                googleServiceAccount: apiGatewaySA.email,  // Critical for logging
            },
        },
        openapiDocuments: [{
            document: {
                path: "openapi.yaml",
                contents: pulumi.asset.StringAsset(openapiSpec),
            },
        }],
    });
    
    // Step 4: Enable audit logging
    const auditConfig = new gcp.projects.IAMAuditConfig("api-gateway-audit", {
        project: "my-project",
        service: "apigateway.googleapis.com",
        auditLogConfigs: [
            { logType: "ADMIN_READ" },
            { logType: "DATA_READ" },
            { logType: "DATA_WRITE" },
        ],
    });
    

    app-engine-managed-updates-enabled

    Severity: high · Enforcement: advisory

    Ensure App Engine managed updates are enabled

    • 2.2 — Ensure that only currently supported software is designated as authorized in the software inventory for enterprise assets. If software is unsupported, yet necessary for the fulfillment of the enterprise’s mission, document an exception detailing mitigating controls and residual risk acceptance. For any unsupported software without an exception documentation, designate as unauthorized. Review the software list to verify software support at least monthly, or more frequently.
    Remediation

    Configure App Engine versions with automaticScaling instead of manualScaling or basicScaling to enable managed updates:

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Standard App Version with automatic scaling
    const app = new gcp.appengine.StandardAppVersion("app-auto-scaled", {
      service: "myapp",
      runtime: "nodejs20",
      versionId: "v1",
      entrypoint: { shell: "node ./app.js" },
      deployment: {
        zip: { sourceUrl: "https://storage.googleapis.com/bucket/app.zip" },
      },
      automaticScaling: {
        maxConcurrentRequests: 10,
        minIdleInstances: 1,
        maxIdleInstances: 3,
      },
    });
    
    // NON-COMPLIANT: Manual scaling prevents managed updates
    // manualScaling: { instances: 3 }  // Don't use this
    
    // NON-COMPLIANT: Basic scaling limits managed updates
    // basicScaling: { maxInstances: 5 }  // Don't use this
    

    application-load-balancer-managed-ssl-certificate

    Severity: high · Enforcement: advisory

    Ensure Application Load Balancer uses managed SSL certificates

    • 3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).
    Remediation

    Attach SSL certificates to target HTTPS proxies via sslCertificates property. Create and configure SSL policy with minimum TLS 1.2 using minTlsVersion: TLS_1_2 and profile: MODERN or RESTRICTED for secure cipher suites.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create managed SSL certificate
    const managedCert = new gcp.compute.ManagedSslCertificate("ssl-cert", {
        name: "my-managed-cert",
        managed: {
            domains: ["example.com", "www.example.com"]
        }
    });
    
    // Create SSL policy with secure TLS settings
    const sslPolicy = new gcp.compute.SSLPolicy("ssl-policy", {
        name: "secure-ssl-policy",
        profile: "MODERN",
        minTlsVersion: "TLS_1_2"
    });
    
    // Create URL map
    const urlMap = new gcp.compute.URLMap("url-map", {
        name: "app-url-map",
        defaultService: backendService.id
    });
    
    // Create HTTPS target proxy with managed certificate and SSL policy
    const httpsProxy = new gcp.compute.TargetHttpsProxy("https-proxy", {
        name: "app-https-proxy",
        urlMap: urlMap.id,
        sslCertificates: [managedCert.id],
        sslPolicy: sslPolicy.id
    });
    
    // Create forwarding rule
    const forwardingRule = new gcp.compute.GlobalForwardingRule("https-rule", {
        name: "https-forwarding-rule",
        target: httpsProxy.id,
        portRange: "443",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL_MANAGED"
    });
    
    // Regional example
    const regionalCert = new gcp.compute.RegionSslCertificate("regional-cert", {
        name: "regional-ssl-cert",
        region: "us-central1",
        certificate: std.file({ input: "path/to/cert.pem" }).result,
        privateKey: std.file({ input: "path/to/key.pem" }).result
    });
    
    const regionalHttpsProxy = new gcp.compute.RegionTargetHttpsProxy("regional-proxy", {
        name: "regional-https-proxy",
        region: "us-central1",
        urlMap: regionalUrlMap.id,
        sslCertificates: [regionalCert.id],
        sslPolicy: sslPolicy.id
    });
    

    bigquery-audit-logging-enabled

    Severity: high · Enforcement: advisory

    Enable BigQuery audit logging for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Configure projects.IAMAuditConfig resource with service: "bigquery.googleapis.com" and enable DATA_READ and DATA_WRITE log types in auditLogConfigs. Create a log sink (ProjectSink or OrganizationSink) with filter protoPayload.serviceName="bigquery.googleapis.com" to export audit logs for long-term retention.

    Example TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bigQueryAuditConfig = new gcp.projects.IAMAuditConfig("bigquery-audit", {
        project: "your-project-id",
        service: "bigquery.googleapis.com",
        auditLogConfigs: [
            { logType: "DATA_READ" },
            { logType: "DATA_WRITE" },
            { logType: "ADMIN_READ" }
        ]
    });
    
    const bigQueryLogSink = new gcp.logging.ProjectSink("bigquery-logs", {
        destination: "storage.googleapis.com/my-audit-logs-bucket",
        filter: 'protoPayload.serviceName="bigquery.googleapis.com"',
        uniqueWriterIdentity: true
    });
    

    bigquery-dataset-encryption-enabled

    Severity: high · Enforcement: advisory

    Enable BigQuery dataset encryption

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure defaultEncryptionConfiguration.kmsKeyName for datasets or encryptionConfiguration.kmsKeyName for tables. Create a Cloud KMS key in the same region as the BigQuery resource and grant the BigQuery service account encrypt/decrypt permissions on the key.

    Example TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    // Create KMS key in same region as dataset
    const keyRing = new gcp.kms.KeyRing("bigquery-keyring", {
        location: "us"
    });
    
    const cryptoKey = new gcp.kms.CryptoKey("bigquery-key", {
        keyRing: keyRing.id,
        rotationPeriod: "7776000s" // 90 days
    });
    
    // Dataset with CMEK encryption
    const dataset = new gcp.bigquery.Dataset("encrypted-dataset", {
        datasetId: "my_dataset",
        location: "US",
        defaultEncryptionConfiguration: {
            kmsKeyName: cryptoKey.id
        }
    });
    
    // Table with CMEK encryption
    const table = new gcp.bigquery.Table("encrypted-table", {
        datasetId: dataset.datasetId,
        tableId: "my_table",
        encryptionConfiguration: {
            kmsKeyName: cryptoKey.id
        }
    });
    

    bigquery-dataset-public-access-check

    Severity: high · Enforcement: advisory

    Ensure BigQuery datasets are not publicly accessible

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove public access from BigQuery datasets by ensuring IAM policies do not grant allUsers or allAuthenticatedUsers permissions. Review dataset accesses configuration and IAM bindings/members. Grant access only to specific service accounts, users, or groups based on least privilege principles.

    Example TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const dataset = new gcp.bigquery.Dataset("secure-dataset", {
        datasetId: "my_dataset",
        location: "US",
        accesses: [
            {
                role: "READER",
                userByEmail: "analyst@example.com"
            },
            {
                role: "roles/bigquery.dataEditor",
                userByEmail: serviceAccount.email
            }
            // Do NOT include allUsers or allAuthenticatedUsers
        ]
    });
    
    // Using IAM binding for specific members only
    const datasetViewer = new gcp.bigquery.DatasetIamBinding("dataset-viewer", {
        datasetId: dataset.datasetId,
        role: "roles/bigquery.dataViewer",
        members: [
            "user:jane@example.com",
            "serviceAccount:myapp@project.iam.gserviceaccount.com"
            // Do NOT use allUsers or allAuthenticatedUsers
        ]
    });
    

    bigquery-maintenance-settings-check

    Severity: medium · Enforcement: advisory

    Ensure BigQuery maintenance settings are configured

    • 4.1 — Establish and maintain a documented secure configuration process for enterprise assets (end-user devices, including portable and mobile, non-computing/IoT devices, and servers) and software (operating systems and applications). Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Configure BigQuery datasets with appropriate maintenance settings:

    import * as gcp from "@pulumi/gcp";
    
    const dataset = new gcp.bigquery.Dataset("my-dataset", {
      datasetId: "my_dataset",
      description: "Dataset purpose and maintenance requirements",
      defaultTableExpirationMs: 7776000000, // 90 days
      defaultPartitionExpirationMs: 7776000000,
      labels: {
        environment: "production",
        dataRetention: "90days",
      },
    });
    

    certificate-manager-certificate-lifecycle-management

    Severity: medium · Enforcement: advisory

    Ensure proper certificate lifecycle management to prevent expiration

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Use managed certificates for automatic renewal to prevent expiration-related outages. Managed certificates handle renewal automatically through DNS or Load Balancer authorization.

    import * as gcp from "@pulumi/gcp";
    
    // RECOMMENDED: Managed certificate with automatic renewal via DNS
    const managedDnsCert = new gcp.certificatemanager.Certificate("managed-dns-cert", {
        name: "managed-dns-cert",
        description: "Auto-renewed certificate via DNS authorization",
        managed: {
            domains: ["example.com", "www.example.com"],
            dnsAuthorizations: [dnsAuthResource.id]
        }
    });
    
    // RECOMMENDED: Managed certificate with automatic renewal via Load Balancer
    const managedLbCert = new gcp.certificatemanager.Certificate("managed-lb-cert", {
        name: "managed-lb-cert",
        description: "Auto-renewed certificate via Load Balancer",
        managed: {
            domains: ["api.example.com"]
            // No dnsAuthorizations needed for Load Balancer authorization
        }
    });
    
    // NOT RECOMMENDED: Self-managed certificate requires manual renewal
    const selfManagedCert = new gcp.certificatemanager.Certificate("self-managed-cert", {
        name: "custom-cert",
        selfManaged: {
            pemCertificate: certPem,
            pemPrivateKey: keyPem
        }
    });
    

    cloud-armor-logging-enabled

    Severity: high · Enforcement: advisory

    Ensure backend services with Cloud Armor have logging enabled

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Enable logging on backend services that use Cloud Armor security policies. Cloud Armor events are logged through the backend service’s access logs, not the security policy itself.

    import * as gcp from "@pulumi/gcp";
    
    // Step 1: Create Cloud Armor security policy
    const armorPolicy = new gcp.compute.SecurityPolicy("armor-policy", {
        name: "security-policy",
        type: "CLOUD_ARMOR",
        rules: [{
            action: "deny(403)",
            priority: 1000,
            match: {
                versionedExpr: "SRC_IPS_V1",
                config: {
                    srcIpRanges: ["192.168.1.0/24"]
                }
            }
        }]
    });
    
    // Step 2: CRITICAL - Enable logging on backend service
    const backendService = new gcp.compute.BackendService("backend", {
        name: "backend-service",
        securityPolicy: armorPolicy.selfLink,
        logConfig: {
            enable: true,        // REQUIRED for Cloud Armor logging
            sampleRate: 1.0      // 1.0 = 100% logging, 0.5 = 50% sampling
        }
    });
    
    // Regional backend service also needs logging
    const regionalBackend = new gcp.compute.RegionBackendService("regional-backend", {
        name: "regional-backend",
        region: "us-central1",
        securityPolicy: armorPolicy.selfLink,
        logConfig: {
            enable: true,        // REQUIRED for Cloud Armor logging
            sampleRate: 1.0
        }
    });
    

    Cloud Armor logs include:

    • Security policy name
    • Matched rule priority and action
    • Request details and outcomes
    • Attack signatures and threat indicators

    cloud-armor-security-policy-logging-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Armor security policies have logging enabled through their backend services

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Enable logging on backend services that use Cloud Armor security policies. Cloud Armor events are logged through the backend service’s access logs, not the security policy itself.

    import * as gcp from "@pulumi/gcp";
    
    // Step 1: Create Cloud Armor security policy
    const securityPolicy = new gcp.compute.SecurityPolicy("security-policy", {
        name: "my-security-policy",
        rules: [{
            action: "deny(403)",
            priority: 1000,
            match: {
                versionedExpr: "SRC_IPS_V1",
                config: {
                    srcIpRanges: ["192.168.1.0/24"]
                }
            }
        }]
    });
    
    // Step 2: CRITICAL - Attach to backend service with logging enabled
    const backend = new gcp.compute.BackendService("backend", {
        name: "backend-service",
        securityPolicy: securityPolicy.selfLink,
        logConfig: {
            enable: true,        // REQUIRED for Cloud Armor logging
            sampleRate: 1.0      // 1.0 = 100% logging, 0.5 = 50% sampling
        }
    });
    
    // Regional backend service also needs logging
    const regionalBackend = new gcp.compute.RegionBackendService("regional-backend", {
        name: "regional-backend",
        region: "us-central1",
        securityPolicy: securityPolicy.selfLink,
        logConfig: {
            enable: true,
            sampleRate: 1.0
        }
    });
    

    Cloud Armor logs include:

    • Security policy name and evaluation results
    • Matched rule priority and action taken
    • Request details and outcomes
    • Attack signatures and threat indicators

    cloud-audit-logs-cloud-logging-enabled

    Severity: high · Enforcement: advisory

    Enable Cloud Audit Logs Cloud Logging integration for monitoring

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Create log sinks to route audit logs to Cloud Logging destinations. Configure gcp.logging.ProjectSink, gcp.logging.OrganizationSink, or gcp.logging.FolderSink resources with filters to capture audit logs (e.g., filter including ‘cloudaudit.googleapis.com’). Route logs to Cloud Logging buckets, BigQuery, or Pub/Sub for centralized analysis and retention.

    cloud-audit-logs-data-access-events-enabled

    Severity: high · Enforcement: advisory

    Enable Cloud Audit Logs data access events for monitoring

    • 8.5 — Configure detailed audit logging for enterprise assets containing sensitive data. Include event source, date, username, timestamp, source addresses, destination addresses, and other useful elements that could assist in a forensic investigation.
    Remediation

    Configure projects.IAMAuditConfig resources with auditLogConfigs including DATA_READ and DATA_WRITE log types for services handling sensitive data to monitor unauthorized access.

    Example - Enable Data Access Audit Logging:

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Enable data access logging for all services
    const auditConfigAll = new gcp.projects.IAMAuditConfig("audit-all-services", {
        project: "my-project-id",
        service: "allServices",  // Apply to all GCP services
        auditLogConfigs: [
            {
                logType: "ADMIN_READ"  // Administrative read operations
            },
            {
                logType: "DATA_READ",  // Compliant: Data read logging enabled
                exemptedMembers: []  // No exemptions
            },
            {
                logType: "DATA_WRITE"  // Compliant: Data write logging enabled
            }
        ]
    });
    
    // COMPLIANT: Enable data access logging for specific services
    const auditConfigStorage = new gcp.projects.IAMAuditConfig("audit-storage", {
        project: "my-project-id",
        service: "storage.googleapis.com",
        auditLogConfigs: [
            {
                logType: "DATA_READ",  // Compliant: Monitor bucket/object reads
                exemptedMembers: ["user:admin@example.com"]  // Optional exemptions
            },
            {
                logType: "DATA_WRITE"  // Compliant: Monitor bucket/object writes
            }
        ]
    });
    
    const auditConfigBigQuery = new gcp.projects.IAMAuditConfig("audit-bigquery", {
        project: "my-project-id",
        service: "bigquery.googleapis.com",
        auditLogConfigs: [
            {
                logType: "DATA_READ"   // Compliant: Monitor query operations
            },
            {
                logType: "DATA_WRITE"  // Compliant: Monitor data modifications
            }
        ]
    });
    
    // NON-COMPLIANT: Only ADMIN_READ enabled, missing data access logging
    const auditConfigIncomplete = new gcp.projects.IAMAuditConfig("audit-incomplete", {
        project: "my-project-id",
        service: "allServices",
        auditLogConfigs: [
            {
                logType: "ADMIN_READ"  // Only admin operations logged
                // Missing DATA_READ and DATA_WRITE
            }
        ]
    });
    
    // NON-COMPLIANT: No audit configuration at all
    // (missing IAMAuditConfig resource)
    

    cloud-audit-logs-integrity-monitoring-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Audit Logs integrity monitoring is enabled

    • 8.3 — Ensure that logging destinations maintain adequate storage to comply with the enterprise’s audit log management process.
    Remediation

    Configure log buckets with locked retention policies to ensure audit log integrity. Set retentionDays to at least 90 days and enable locked: true on gcp.logging.LogBucket resources. Locked retention policies prevent modification or deletion, ensuring audit logs remain trustworthy. Route audit logs to BigQuery or Cloud Storage for tamper-evident long-term storage.

    cloud-audit-logs-multi-region-trail-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Audit Logs multi-region trail is enabled

    • 8.3 — Ensure that logging destinations maintain adequate storage to comply with the enterprise’s audit log management process.
    Remediation

    Configure Cloud Audit Logs to export to a multi-region or dual-region Cloud Storage bucket. Update the logging sink destination to use a bucket with location set to ‘US’, ‘EU’, ‘ASIA’, or a dual-region location for high availability.

    cloud-audit-logs-security-trail-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Audit Logs security trail is enabled

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Enable Cloud Audit Logs by creating a gcp.logging.ProjectSink or gcp.logging.OrganizationSink resource that captures admin activity and data access logs. Configure the sink to export logs to Cloud Storage, BigQuery, or Pub/Sub for long-term retention and analysis.

    cloud-build-trigger-envvar-gcpcred-check

    Severity: high · Enforcement: advisory

    Ensure Cloud Build trigger environment variables do not contain GCP credentials

    • 16.1 — Establish and maintain a secure application development process. In the process, address such items as: secure application design standards, secure coding practices, developer training, vulnerability management, security of third-party code, and application security testing procedures. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    • 16.12 — Apply static and dynamic analysis tools within the application life cycle to verify that secure coding practices are being followed.
    Remediation

    Remove GCP credentials from Cloud Build trigger substitutions, build.options.envs, and step environment variables. Store credentials in Secret Manager and reference them using availableSecrets configuration. Grant Cloud Build service account access to the secrets.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Using Secret Manager references
    const compliantTrigger = new gcp.cloudbuild.Trigger("compliant-trigger", {
        location: "us-central1",
        triggerTemplate: {
            branchName: "main",
            repoName: "my-repo",
        },
        substitutions: {
            _ENV: "production",
            _REGION: "us-central1",
        },
        filename: "cloudbuild.yaml",
    });
    
    // NON-COMPLIANT: Storing credentials in substitutions
    const badTrigger = new gcp.cloudbuild.Trigger("bad-trigger", {
        location: "us-central1",
        triggerTemplate: {
            branchName: "main",
            repoName: "my-repo",
        },
        substitutions: {
            _GCP_API_KEY: "AIzaSyD...", // Violates policy
            _SERVICE_ACCOUNT_KEY: '{"type":"service_account",...}', // Violates policy
        },
        filename: "cloudbuild.yaml",
    });
    
    // NON-COMPLIANT: Storing credentials in build options
    const badTriggerWithEnvs = new gcp.cloudbuild.Trigger("bad-trigger-envs", {
        name: "my-trigger",
        build: {
            options: {
                envs: [
                    "ENV = production",
                    "GCP_CREDENTIALS = {...}", // Violates policy
                ],
            },
            steps: [{
                name: "gcr.io/cloud-builders/gcloud",
                args: ["version"],
            }],
        },
    });
    

    cloud-build-trigger-source-repo-url-check

    Severity: medium · Enforcement: advisory

    Ensure Cloud Build trigger source repository URLs use secure protocols

    • 16.1 — Establish and maintain a secure application development process. In the process, address such items as: secure application design standards, secure coding practices, developer training, vulnerability management, security of third-party code, and application security testing procedures. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    • 16.12 — Apply static and dynamic analysis tools within the application life cycle to verify that secure coding practices are being followed.
    Remediation

    Update source repository URLs in sourceToBuild.uri, gitFileSource.uri, and other URL fields to use HTTPS protocol. Replace http://, git://, or ftp:// protocols with https:// or ssh:// (git@). Configure proper authentication for the repository connections.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Using HTTPS protocol
    const compliantTrigger = new gcp.cloudbuild.Trigger("compliant-trigger", {
        name: "my-trigger",
        location: "us-central1",
        sourceToBuild: {
            uri: "https://github.com/my-org/my-repo", // Secure HTTPS
            ref: "refs/heads/main",
            repoType: "GITHUB",
        },
        filename: "cloudbuild.yaml",
    });
    
    // COMPLIANT: Using SSH protocol
    const compliantSshTrigger = new gcp.cloudbuild.Trigger("compliant-ssh-trigger", {
        name: "my-ssh-trigger",
        location: "us-central1",
        sourceToBuild: {
            uri: "git@github.com:my-org/my-repo.git", // Secure SSH
            ref: "refs/heads/main",
            repoType: "GITHUB",
        },
        filename: "cloudbuild.yaml",
    });
    
    // NON-COMPLIANT: Using insecure HTTP protocol
    const badTrigger = new gcp.cloudbuild.Trigger("bad-trigger", {
        name: "bad-trigger",
        location: "us-central1",
        sourceToBuild: {
            uri: "http://github.com/my-org/my-repo", // Violates policy - insecure HTTP
            ref: "refs/heads/main",
            repoType: "GITHUB",
        },
        filename: "cloudbuild.yaml",
    });
    
    // NON-COMPLIANT: Using insecure git:// protocol
    const badGitTrigger = new gcp.cloudbuild.Trigger("bad-git-trigger", {
        name: "bad-git-trigger",
        location: "us-central1",
        sourceToBuild: {
            uri: "git://github.com/my-org/my-repo.git", // Violates policy - insecure git://
            ref: "refs/heads/main",
            repoType: "GITHUB",
        },
        filename: "cloudbuild.yaml",
    });
    

    cloud-cdn-logging-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud CDN has logging enabled for audit and monitoring

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Configure Backend Services with logConfig.enable: true when enableCdn: true. Set an appropriate sampleRate (0.0 to 1.0) to balance log volume with coverage. Logs are automatically sent to Cloud Logging for centralized management.

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    // Create a backend service with Cloud CDN enabled and logging configured
    const backendService = new gcp.compute.BackendService("my-backend", {
        name: "my-backend-service",
        protocol: "HTTP",
        timeoutSec: 30,
        // Enable Cloud CDN
        enableCdn: true,
        // Configure CDN policy
        cdnPolicy: {
            cacheMode: "CACHE_ALL_STATIC",
            defaultTtl: 3600,
            clientTtl: 7200,
            maxTtl: 86400,
            negativeCaching: true,
        },
        // Enable logging for Cloud CDN
        logConfig: {
            enable: true,
            sampleRate: 1.0, // Log 100% of requests (adjust based on traffic volume)
        },
        backends: [{
            group: instanceGroup.id,
            balancingMode: "UTILIZATION",
            capacityScaler: 1.0,
        }],
        healthChecks: [healthCheck.id],
    });
    

    cloud-filestore-encrypted-check

    Severity: high · Enforcement: advisory

    Ensure Cloud Filestore is encrypted

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure customer-managed encryption keys (CMEK) by setting kmsKeyName to a Cloud KMS key when creating Filestore instances for enhanced security.

    import * as gcp from "@pulumi/gcp";
    
    // Create KMS key for encryption
    const filestoreKeyring = new gcp.kms.KeyRing("filestore-keyring", {
        name: "filestore-keyring",
        location: "us-central1"
    });
    
    const filestoreKey = new gcp.kms.CryptoKey("filestore-key", {
        name: "filestore-key",
        keyRing: filestoreKeyring.id
    });
    
    // COMPLIANT: Filestore instance with CMEK configured
    const instanceCompliant = new gcp.filestore.Instance("encrypted-filestore", {
        name: "encrypted-filestore",
        location: "us-central1",
        tier: "ENTERPRISE",
        fileShares: {
            capacityGb: 1024,
            name: "share1"
        },
        networks: [{
            network: "default",
            modes: ["MODE_IPV4"]
        }],
        kmsKeyName: filestoreKey.id  // Compliant: CMEK configured
    });
    
    // NON-COMPLIANT: Filestore instance without CMEK
    const instanceNonCompliant = new gcp.filestore.Instance("default-encrypted", {
        name: "default-encrypted",
        location: "us-central1",
        tier: "BASIC_HDD",
        fileShares: {
            capacityGb: 1024,
            name: "share1"
        },
        networks: [{
            network: "default",
            modes: ["MODE_IPV4"]
        }]
        // Missing kmsKeyName - uses Google-managed keys
    });
    

    cloud-firestore-autoscaling-enabled

    Severity: medium · Enforcement: advisory

    Ensure Cloud Firestore autoscaling is enabled

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Configure Cloud Firestore in Native mode (type: FIRESTORE_NATIVE) for automatic scaling. Use OPTIMISTIC concurrency mode for better scalability and performance.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Firestore Native mode with OPTIMISTIC concurrency (automatic scaling)
    const databaseCompliant = new gcp.firestore.Database("native-optimistic", {
        name: "production-db",
        locationId: "nam5",  // Multi-region for better availability
        type: "FIRESTORE_NATIVE",  // Compliant: has automatic scaling
        concurrencyMode: "OPTIMISTIC",  // Compliant: best for scalability
        appEngineIntegrationMode: "DISABLED",
        deleteProtectionState: "DELETE_PROTECTION_ENABLED",
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_ENABLED"
    });
    
    // NON-COMPLIANT: Firestore Native with PESSIMISTIC concurrency
    const databasePessimistic = new gcp.firestore.Database("native-pessimistic", {
        name: "db-pessimistic",
        locationId: "us-central1",
        type: "FIRESTORE_NATIVE",
        concurrencyMode: "PESSIMISTIC",  // Non-compliant: can limit throughput
        appEngineIntegrationMode: "DISABLED"
    });
    
    // NON-COMPLIANT: Datastore mode instead of Native mode
    const databaseDatastore = new gcp.firestore.Database("datastore-mode", {
        name: "datastore-db",
        locationId: "us-central1",
        type: "DATASTORE_MODE"  // Non-compliant: Native mode recommended for enhanced scaling
    });
    
    // NON-COMPLIANT: Single-region without multi-region availability
    const databaseSingleRegion = new gcp.firestore.Database("single-region", {
        name: "single-region-db",
        locationId: "us-west1",  // Non-compliant: not multi-region
        type: "FIRESTORE_NATIVE"
    });
    

    cloud-firestore-in-backup-plan

    Severity: high · Enforcement: advisory

    Perform automated backups for Cloud Firestore databases and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Create backup schedules with appropriate retention periods. Enable point-in-time recovery by setting pointInTimeRecoveryEnablement to ‘POINT_IN_TIME_RECOVERY_ENABLED’.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Firestore database with PITR and backup schedule
    const database = new gcp.firestore.Database("production-db", {
        name: "production-db",
        locationId: "nam5",
        type: "FIRESTORE_NATIVE",
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_ENABLED",  // Compliant
        deleteProtectionState: "DELETE_PROTECTION_ENABLED"
    });
    
    const backupSchedule = new gcp.firestore.BackupSchedule("daily-backup", {
        database: database.name,
        retention: "30d",  // Compliant: adequate retention period
        dailyRecurrence: {}  // Compliant: daily backups configured
    });
    
    // NON-COMPLIANT: Database without backup schedule
    const databaseNoBackup = new gcp.firestore.Database("no-backup-db", {
        name: "test-db",
        locationId: "us-central1",
        type: "FIRESTORE_NATIVE",
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_DISABLED"  // Non-compliant
        // Missing backup schedule
    });
    
    // NON-COMPLIANT: Backup schedule with insufficient retention
    const shortRetentionBackup = new gcp.firestore.BackupSchedule("short-backup", {
        database: database.name,
        retention: "3d",  // Non-compliant: less than 7 days
        dailyRecurrence: {}
    });
    
    // NON-COMPLIANT: Backup schedule without recurrence
    const noRecurrenceBackup = new gcp.firestore.BackupSchedule("no-recurrence", {
        database: database.name,
        retention: "30d"
        // Missing dailyRecurrence or weeklyRecurrence
    });
    

    cloud-firestore-pitr-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Firestore point-in-time recovery is enabled and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Enable point-in-time recovery by setting pointInTimeRecoveryEnablement to ‘POINT_IN_TIME_RECOVERY_ENABLED’. Also set deleteProtectionState to ‘DELETE_PROTECTION_ENABLED’.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Firestore database with PITR and deletion protection enabled
    const databaseCompliant = new gcp.firestore.Database("compliant-db", {
        name: "production-db",
        locationId: "nam5",
        type: "FIRESTORE_NATIVE",
        concurrencyMode: "OPTIMISTIC",
        appEngineIntegrationMode: "DISABLED",
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_ENABLED",  // Compliant
        deleteProtectionState: "DELETE_PROTECTION_ENABLED"  // Compliant
    });
    
    // NON-COMPLIANT: Firestore database without PITR
    const databaseNonCompliant = new gcp.firestore.Database("no-pitr-db", {
        name: "test-db",
        locationId: "us-central1",
        type: "FIRESTORE_NATIVE",
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_DISABLED",  // Non-compliant
        deleteProtectionState: "DELETE_PROTECTION_DISABLED"  // Non-compliant
    });
    
    // NON-COMPLIANT: Datastore mode (PITR not available)
    const datastoreMode = new gcp.firestore.Database("datastore-db", {
        name: "datastore-db",
        locationId: "us-central1",
        type: "DATASTORE_MODE",  // Non-compliant: PITR only available in FIRESTORE_NATIVE
        pointInTimeRecoveryEnablement: "POINT_IN_TIME_RECOVERY_ENABLED"
    });
    

    cloud-functions-concurrency-check

    Severity: medium · Enforcement: advisory

    Configure Cloud Functions concurrency for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    For v1 functions, set maxInstances to limit concurrent executions. For v2 functions, configure serviceConfig.maxInstanceRequestConcurrency (1 for CPU-bound, 80+ for I/O-bound) and serviceConfig.maxInstanceCount. Base values on expected load and resource requirements for proper capacity monitoring.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: 1st gen function with maxInstances configured
    const functionV1 = new gcp.cloudfunctions.Function("my-function-v1", {
        name: "example-function",
        runtime: "nodejs20",
        sourceArchiveBucket: bucket.name,
        sourceArchiveObject: archive.name,
        entryPoint: "helloWorld",
        triggerHttp: true,
        maxInstances: 10, // Limits concurrent instances
    });
    
    // COMPLIANT: 2nd gen function with concurrency settings
    const functionV2 = new gcp.cloudfunctionsv2.Function("my-function-v2", {
        name: "example-function-v2",
        location: "us-central1",
        buildConfig: {
            runtime: "nodejs20",
            entryPoint: "helloWorld",
            source: {
                storageSource: {
                    bucket: bucket.name,
                    object: archive.name,
                },
            },
        },
        serviceConfig: {
            maxInstanceCount: 10,
            maxInstanceRequestConcurrency: 80, // For I/O-bound functions
        },
    });
    
    // NON-COMPLIANT: Missing maxInstances
    const badFunction = new gcp.cloudfunctions.Function("bad-function", {
        name: "bad-example",
        runtime: "nodejs20",
        triggerHttp: true,
        // maxInstances not configured - violates policy
    });
    

    cloud-functions-inside-vpc

    Severity: high · Enforcement: advisory

    Ensure Cloud Functions are inside VPC

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Configure VPC connector for Cloud Functions:

    // Cloud Functions v1
    new gcp.cloudfunctions.Function("my-function", {
      vpcConnector: "projects/my-project/locations/us-central1/connectors/my-connector",
      vpcConnectorEgressSettings: "ALL_TRAFFIC", // or "PRIVATE_RANGES_ONLY"
    });
    
    // Cloud Functions v2
    new gcp.cloudfunctionsv2.Function("my-function-v2", {
      serviceConfig: {
        vpcConnector: "projects/my-project/locations/us-central1/connectors/my-connector",
        vpcConnectorEgressSettings: "ALL_TRAFFIC",
      },
    });
    

    cloud-functions-public-access-prohibited

    Severity: high · Enforcement: advisory

    Ensure Cloud Functions prohibit public access

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM bindings and members that grant allUsers or allAuthenticatedUsers permissions. Use Cloud IAM to grant roles/cloudfunctions.invoker only to specific service accounts, users, or groups. For HTTP-triggered functions requiring public access, implement authentication in the function code.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Grant access to specific user
    const member = new gcp.cloudfunctions.FunctionIamMember("function-member", {
        project: myFunction.project,
        region: myFunction.region,
        cloudFunction: myFunction.name,
        role: "roles/cloudfunctions.invoker",
        member: "user:jane@example.com",
    });
    
    // NON-COMPLIANT: Grants public access
    const publicMember = new gcp.cloudfunctions.FunctionIamMember("public-member", {
        cloudFunction: myFunction.name,
        role: "roles/cloudfunctions.invoker",
        member: "allUsers", // Violates policy
    });
    

    cloud-logging-encrypted

    Severity: high · Enforcement: advisory

    Ensure Cloud Logging is encrypted

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure Cloud Logging buckets with customer-managed encryption keys (CMEK) for enhanced security. On gcp.logging.LogBucket or gcp.logging.ProjectBucketConfig resources, set cmekSettings with a kmsKeyName pointing to a Cloud KMS key. While GCP encrypts all log data by default with Google-managed keys, CMEK provides additional control and compliance capabilities.

    cloud-logging-retention-period-365

    Severity: high · Enforcement: advisory

    Ensure Cloud Logging log retention is 365 days or more

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Configure Cloud Logging buckets with retentionDays set to at least 365 days. Update gcp.logging.ProjectBucketConfig, gcp.logging.OrganizationBucketConfig, gcp.logging.FolderBucketConfig, or gcp.logging.BillingAccountBucketConfig resources to specify ‘retentionDays: 365’ or higher to maintain adequate audit trails for security investigations and compliance.

    cloud-run-service-configuration-check

    Severity: high · Enforcement: advisory

    Ensure Cloud Run service configuration check

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM bindings that grant allUsers the roles/run.invoker role. Set service ingress annotation run.googleapis.com/ingress to internal or internal-and-cloud-load-balancing in metadata.annotations. Grant invocation permissions only to specific service accounts or users.

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Service with restricted ingress
    const compliantService = new gcp.cloudrun.Service("compliant-service", {
        name: "secure-service",
        location: "us-central1",
        metadata: {
            annotations: {
                "run.googleapis.com/ingress": "internal-and-cloud-load-balancing",
            },
        },
        template: {
            spec: {
                containers: [{
                    image: "gcr.io/my-project/my-image:latest",
                }],
            },
        },
    });
    
    // COMPLIANT: Grant access to specific user
    const specificAccess = new gcp.cloudrun.IamMember("specific-access", {
        service: compliantService.name,
        location: compliantService.location,
        role: "roles/run.invoker",
        member: "user:jane@example.com",
    });
    
    // NON-COMPLIANT: Public access via IAM
    const publicAccess = new gcp.cloudrun.IamMember("public-access", {
        service: compliantService.name,
        location: compliantService.location,
        role: "roles/run.invoker",
        member: "allUsers", // Violates policy
    });
    
    // NON-COMPLIANT: Service without ingress restrictions
    const badService = new gcp.cloudrun.Service("bad-service", {
        name: "public-service",
        location: "us-central1",
        metadata: {
            annotations: {
                "run.googleapis.com/ingress": "all", // Violates policy
            },
        },
        template: {
            spec: {
                containers: [{
                    image: "gcr.io/my-project/my-image:latest",
                }],
            },
        },
    });
    

    cloud-security-scanner-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Security Scanner is enabled for vulnerability management

    • 7.1 — Establish and maintain a documented vulnerability management process for enterprise assets. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Enable the Web Security Scanner API service (websecurityscanner.googleapis.com) in your project. Configure scan configurations for your web applications using gcp.securityscanner.ScanConfig resources. Specify starting URLs, authentication details, and scan schedules to ensure comprehensive vulnerability coverage.

    Example - Enable Web Security Scanner:

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Enable Web Security Scanner API and configure scan
    const scannerService = new gcp.projects.Service("scanner-api", {
        service: "websecurityscanner.googleapis.com"  // Compliant: API enabled
    });
    
    const scanConfig = new gcp.securityscanner.ScanConfig("web-scan", {
        displayName: "Production Web App Scan",
        startingUrls: ["https://myapp.example.com"],  // Compliant: URLs configured
        schedule: {
            scheduleTime: "0 2 * * *",  // Compliant: daily at 2 AM
            intervalDurationDays: 1
        },
        maxQps: 5,
        targetPlatforms: ["APP_ENGINE", "COMPUTE"]
    }, { dependsOn: [scannerService] });
    
    // NON-COMPLIANT: Web Security Scanner API not enabled
    // (no gcp.projects.Service resource for websecurityscanner.googleapis.com)
    
    // NON-COMPLIANT: Scanner API enabled but no scan configurations
    const scannerServiceOnly = new gcp.projects.Service("scanner-only", {
        service: "websecurityscanner.googleapis.com"
        // Missing ScanConfig resources
    });
    
    // NON-COMPLIANT: Scan config without schedule
    const scanConfigNoSchedule = new gcp.securityscanner.ScanConfig("no-schedule-scan", {
        displayName: "Manual Scan Only",
        startingUrls: ["https://app.example.com"]
        // Missing schedule - won't run automatically
    });
    
    // NON-COMPLIANT: Scan config without starting URLs
    const scanConfigNoUrls = new gcp.securityscanner.ScanConfig("no-urls-scan", {
        displayName: "Incomplete Scan",
        schedule: {
            scheduleTime: "0 2 * * *"
        }
        // Missing startingUrls - scanner can't run
    });
    

    cloud-sql-backup-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud SQL backup is enabled and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Enable automated backups and deletion protection. Set backupConfiguration.enabled to true, deletionProtection to true, and configure cross-region backup location for isolated recovery data.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("my-db-instance", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        deletionProtection: true,
        settings: {
            tier: "db-f1-micro",
            availabilityType: "REGIONAL",
            backupConfiguration: {
                enabled: true,
                startTime: "03:00",
                pointInTimeRecoveryEnabled: true,
                location: "us-west1", // Different region for isolation
                backupRetentionSettings: {
                    retainedBackups: 7,
                    retentionUnit: "COUNT",
                },
                transactionLogRetentionDays: 7,
            },
        },
    });
    

    cloud-sql-enhanced-networking-enabled

    Severity: medium · Enforcement: advisory

    Ensure Cloud SQL enhanced networking is enabled

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Enable private IP by setting ipv4Enabled to false and configuring privateNetwork. Set requireSsl to true to enforce encrypted connections.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("secure-db", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        settings: {
            tier: "db-f1-micro",
            ipConfiguration: {
                ipv4Enabled: false,
                privateNetwork: vpc.id,
                requireSsl: true,
            },
        },
    });
    

    cloud-sql-instance-backup-enabled

    Severity: high · Enforcement: advisory

    Perform automated backups for Cloud SQL instances

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    Remediation

    Enable automated backups by setting backupConfiguration.enabled to true. Configure startTime, pointInTimeRecoveryEnabled, and backupRetentionSettings with at least 7 days retention.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("my-db-instance", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        settings: {
            tier: "db-f1-micro",
            backupConfiguration: {
                enabled: true,
                startTime: "03:00",
                pointInTimeRecoveryEnabled: true,
                backupRetentionSettings: {
                    retainedBackups: 7,
                    retentionUnit: "COUNT",
                },
            },
        },
    });
    

    cloud-sql-instance-deletion-protection-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud SQL instance deletion protection is enabled

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Enable deletion protection by setting deletionProtection to true. Also enable automated backups and point-in-time recovery for comprehensive data protection.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("my-db-instance", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        deletionProtection: true,
        settings: {
            tier: "db-f1-micro",
            backupConfiguration: {
                enabled: true,
                pointInTimeRecoveryEnabled: true,
            },
        },
    });
    

    cloud-sql-instance-not-publicly-accessible

    Severity: high · Enforcement: advisory

    Restrict public access for Cloud SQL instances

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Disable public IP by setting ipv4Enabled to false and configure privateNetwork in ipConfiguration. Remove authorized networks that allow 0.0.0.0/0 access.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("my-db-instance", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        settings: {
            tier: "db-f1-micro",
            ipConfiguration: {
                ipv4Enabled: false,
                privateNetwork: "projects/my-project/global/networks/my-vpc",
            },
        },
    });
    

    cloud-sql-logging-enabled

    Severity: high · Enforcement: advisory

    Enable Cloud SQL logging for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Enable database logging flags via databaseFlags. For PostgreSQL: set log_connections, log_disconnections, log_checkpoints to ‘on’. For MySQL: enable general_log or slow_query_log.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("db-with-logging", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        settings: {
            tier: "db-f1-micro",
            databaseFlags: [
                { name: "log_connections", value: "on" },
                { name: "log_disconnections", value: "on" },
                { name: "log_checkpoints", value: "on" },
            ],
        },
    });
    

    cloud-sql-multi-region-support

    Severity: medium · Enforcement: advisory

    Ensure Cloud SQL multi-region support is enabled

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Configure Cloud SQL with REGIONAL availability type for high availability. Deploy cross-region read replicas and configure multi-region backup storage for disaster recovery.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("ha-db", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        settings: {
            tier: "db-f1-micro",
            availabilityType: "REGIONAL",
            backupConfiguration: {
                enabled: true,
                location: "us", // Multi-region backup
            },
        },
    });
    

    cloud-sql-storage-encrypted

    Severity: high · Enforcement: advisory

    Ensure Cloud SQL storage is encrypted

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure customer-managed encryption keys (CMEK) by setting encryptionKeyName to a Cloud KMS key for enhanced security and key management control.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.sql.DatabaseInstance("my-db-instance", {
        databaseVersion: "POSTGRES_15",
        region: "us-central1",
        encryptionKeyName: "projects/my-project/locations/us-central1/keyRings/my-keyring/cryptoKeys/my-key",
        settings: {
            tier: "db-f1-micro",
        },
    });
    

    cloud-storage-bucket-cross-region-replication-enabled

    Severity: high · Enforcement: advisory

    Perform automated backups for Cloud Storage buckets with cross-region replication and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Use multi-region (US, EU, ASIA) or dual-region storage locations for automatic cross-region replication. Enable versioning and configure lifecycle rules for backup management.

    cloud-storage-bucket-default-encryption-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage buckets have default encryption enabled

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Configure Cloud Storage buckets to use customer-managed encryption keys (CMEK) by setting the ’encryption.defaultKmsKeyName’ property with a Cloud KMS key.

    cloud-storage-bucket-encryption-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage bucket encryption is enabled

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure the encryption property with defaultKmsKeyName to specify a Cloud KMS key for bucket encryption using customer-managed encryption keys (CMEK).

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("my-bucket", {
        location: "US",
        encryption: {
            defaultKmsKeyName: "projects/my-project/locations/us/keyRings/my-keyring/cryptoKeys/my-key",
        },
    });
    

    cloud-storage-bucket-iam-policy-grantee-check

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage bucket IAM policy grantee check

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM policy bindings that grant access to allUsers or allAuthenticatedUsers. Use Cloud IAM to grant access only to specific service accounts, users, or groups.

    cloud-storage-bucket-level-public-access-prohibited

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage bucket level public access is prohibited

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Enable uniform bucket-level access and set publicAccessPrevention to ’enforced’ to prevent individual objects from being made public.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("private-bucket", {
        location: "US",
        uniformBucketLevelAccess: true,
        publicAccessPrevention: "enforced",
    });
    

    cloud-storage-bucket-logging-enabled

    Severity: high · Enforcement: advisory

    Enable Cloud Storage bucket logging for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Configure the logging property with logBucket set to a destination bucket and optionally logObjectPrefix for organized log storage.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const logBucket = new gcp.storage.Bucket("log-bucket", {
        location: "US",
    });
    
    const bucket = new gcp.storage.Bucket("monitored-bucket", {
        location: "US",
        logging: {
            logBucket: logBucket.name,
            logObjectPrefix: "access-logs/",
        },
    });
    

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("my-bucket", {
        location: "US",
        logging: {
            logBucket: "my-log-bucket",
            logObjectPrefix: "bucket-logs/",
        },
    });
    

    cloud-storage-bucket-public-read-prohibited

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage bucket public read is prohibited

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove public read access from Cloud Storage buckets by ensuring IAM policies and ACLs do not grant allUsers or allAuthenticatedUsers read permissions. Set ‘publicAccessPrevention’ to ’enforced’ to prevent public access.

    cloud-storage-bucket-public-write-prohibited

    Severity: critical · Enforcement: advisory

    Ensure Cloud Storage bucket public write is prohibited

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM policy bindings that grant write permissions to allUsers or allAuthenticatedUsers. Ensure bucket ACLs do not grant WRITE or FULL_CONTROL to public principals.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("secure-bucket", {
        location: "US",
        uniformBucketLevelAccess: true,
        publicAccessPrevention: "enforced",
    });
    
    // Do NOT add IAM bindings like this:
    // new gcp.storage.BucketIAMMember("bad-binding", {
    //     bucket: bucket.name,
    //     role: "roles/storage.objectCreator",
    //     member: "allUsers",  // NEVER use allUsers/allAuthenticatedUsers
    // });
    

    cloud-storage-bucket-restrict-public-read-access

    Severity: high · Enforcement: advisory

    Restrict public access for Cloud Storage buckets

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM bindings granting allUsers or allAuthenticatedUsers read permissions. Enable uniform bucket-level access and use specific service accounts or groups.

    cloud-storage-bucket-ssl-requests-only

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage buckets require SSL requests only

    • 3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).
    Remediation

    Enable uniform bucket-level access and set publicAccessPrevention to ’enforced’. Configure IAM conditions requiring secure transport for all bucket access.

    cloud-storage-bucket-versioning-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage bucket versioning is enabled

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Enable versioning on Cloud Storage buckets by setting versioning.enabled to true to protect against accidental deletion or modification.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("versioned-bucket", {
        location: "US",
        versioning: {
            enabled: true,
        },
    });
    

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const bucket = new gcp.storage.Bucket("my-bucket", {
        location: "US",
        versioning: {
            enabled: true,
        },
    });
    

    cloud-storage-default-encryption-kms

    Severity: high · Enforcement: advisory

    Ensure Cloud Storage default encryption uses Cloud KMS

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure the encryption property with defaultKmsKeyName to specify a Cloud KMS key for bucket encryption.

    compute-engine-instance-detailed-monitoring-enabled

    Severity: medium · Enforcement: advisory

    Enable Compute Engine instance detailed monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Enable shieldedInstanceConfig with enableIntegrityMonitoring set to true. Add metadata keys ’enable-guest-attributes: TRUE’ and ’enable-oslogin: TRUE’. Install the Ops Agent for comprehensive monitoring.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.compute.Instance("monitored-instance", {
        machineType: "e2-medium",
        zone: "us-central1-a",
        shieldedInstanceConfig: {
            enableIntegrityMonitoring: true,
            enableVtpm: true,
        },
        metadata: {
            "enable-guest-attributes": "TRUE",
            "enable-oslogin": "TRUE",
        },
        bootDisk: { /* ... */ },
        networkInterfaces: [{ /* ... */ }],
    });
    

    compute-engine-instance-in-vpc

    Severity: high · Enforcement: advisory

    Ensure Compute Engine instances are deployed in VPC networks

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Deploy instances in custom VPC with explicit subnetwork:

    new gcp.compute.Instance("my-instance", {
      networkInterfaces: [{
        network: customVpc.id,
        subnetwork: customSubnet.id,
        // Specify explicit VPC and subnet
      }],
      // ... other config
    });
    

    Avoid using legacy default networks.

    compute-engine-instance-not-publicly-accessible

    Severity: high · Enforcement: advisory

    Ensure Compute Engine instances are not publicly accessible

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove accessConfigs from networkInterfaces to disable external IP assignment:

    new gcp.compute.Instance("my-instance", {
      networkInterfaces: [{
        network: "my-vpc",
        subnetwork: "my-subnet",
        // accessConfigs: [], // Remove or set to empty array
      }],
      canIpForward: false,
    });
    

    Use Cloud NAT or load balancers for controlled external access.

    compute-engine-instance-os-config-managed

    Severity: high · Enforcement: advisory

    Ensure Compute Engine instances are managed with OS Config

    • 12.1 — Ensure network infrastructure is kept up-to-date. Example implementations include running the latest stable release of software and/or using currently supported network as a service (NaaS) offerings. Review software versions monthly, or more frequently, to verify software support.
    Remediation

    Enable OS Config management by adding metadata key ’enable-osconfig: TRUE’ to enable centralized patch management, security updates, and configuration compliance.

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.compute.Instance("managed-instance", {
        machineType: "e2-medium",
        zone: "us-central1-a",
        metadata: {
            "enable-osconfig": "TRUE",  // REQUIRED: Enable OS Config management
        },
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
            },
        },
        networkInterfaces: [{
            network: "default",
            accessConfigs: [{}],
        }],
    });
    

    Benefits of OS Config:

    • Automated security patching
    • Vulnerability detection and reporting
    • Configuration drift detection
    • Centralized compliance monitoring

    compute-engine-instance-proper-configuration

    Severity: medium · Enforcement: advisory

    Ensure Compute Engine instances have proper configuration for lifecycle management

    • 1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.
    Remediation

    Configure instances with appropriate settings based on their type:

    For production instances:

    import * as gcp from "@pulumi/gcp";
    
    const productionInstance = new gcp.compute.Instance("prod-instance", {
        machineType: "e2-medium",
        zone: "us-central1-a",
        scheduling: {
            automaticRestart: true,         // Ensure high availability
            onHostMaintenance: "MIGRATE",   // Migrate during maintenance
            preemptible: false,             // Use standard instances for production
        },
        deletionProtection: true,           // Protect production resources
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
                size: 100,
                type: "pd-ssd",             // Use SSD for production
            },
            autoDelete: false,              // Preserve data on instance deletion
        },
        networkInterfaces: [{
            network: "default",
            accessConfigs: [{}],
        }],
    });
    

    For temporary/development instances:

    const tempInstance = new gcp.compute.Instance("temp-instance", {
        machineType: "e2-medium",
        zone: "us-central1-a",
        scheduling: {
            preemptible: true,              // Use preemptible for cost savings
            automaticRestart: false,        // Don't restart when preempted
            onHostMaintenance: "TERMINATE", // Terminate on maintenance
        },
        deletionProtection: false,         // Allow easy cleanup
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
                size: 20,
                type: "pd-standard",        // Use standard disk for dev
            },
            autoDelete: true,               // Clean up disk with instance
        },
        networkInterfaces: [{
            network: "default",
            accessConfigs: [{}],
        }],
    });
    

    compute-engine-instance-service-account-attached

    Severity: high · Enforcement: advisory

    Ensure Compute Engine instances have service accounts attached

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Attach a custom service account with explicit scopes:

    new gcp.compute.Instance("my-instance", {
      serviceAccount: {
        email: customServiceAccount.email,
        scopes: [
          "https://www.googleapis.com/auth/cloud-platform",
        ],
      },
      // ... other config
    });
    

    Avoid using the default compute service account.

    compute-engine-instance-uses-os-login

    Severity: high · Enforcement: advisory

    Ensure Compute Engine instances enforce OS Login

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Add metadata keys to enable OS Login:

    new gcp.compute.Instance("my-instance", {
      metadata: {
        "enable-oslogin": "TRUE",
        "block-project-ssh-keys": "TRUE",
        "enable-oslogin-2fa": "TRUE", // For production instances
      },
      // ... other config
    });
    

    This enables centralized SSH authentication via IAM.

    compute-engine-optimized-instance

    Severity: medium · Enforcement: advisory

    Ensure Compute Engine instances are optimized for backup and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Enable deletion protection and configure boot disk:

    new gcp.compute.Instance("my-instance", {
      deletionProtection: true,
      bootDisk: {
        autoDelete: false,
        initializeParams: {
          type: "pd-ssd", // Use persistent disk type
        },
      },
      labels: {
        "backup": "enabled",
        "critical": "true",
      },
    });
    

    Avoid scratch disks for data requiring backup.

    compute-engine-os-config-compliance-association-compliant

    Severity: high · Enforcement: advisory

    Ensure OS Config managed instances have compliance association

    • 1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.
    Remediation

    Add compliance labels such as ‘patch-group’ or ‘compliance-policy’ to instances with OS Config enabled. Enable OS Config by setting metadata ’enable-osconfig’ to ‘TRUE’ for proper vulnerability scanning and patch management tracking.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const instance = new gcp.compute.Instance("compliant-instance", {
        machineType: "e2-medium",
        zone: "us-central1-a",
        metadata: {
            "enable-osconfig": "TRUE",
        },
        labels: {
            "patch-group": "production",
            "compliance-policy": "cis-benchmark",
        },
        bootDisk: { /* ... */ },
        networkInterfaces: [{ /* ... */ }],
    });
    

    compute-engine-os-config-compliance-patch-compliant

    Severity: high · Enforcement: advisory

    Ensure OS Config managed instances have patch compliance

    • 7.1 — Establish and maintain a documented vulnerability management process for enterprise assets. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    • 7.3 — Perform operating system updates on enterprise assets through automated patch management on a monthly, or more frequent, basis.
    Remediation

    Enable OS Config on instances by setting metadata ’enable-osconfig: TRUE’. Create patch deployment policies with schedules and instance filters. Configure patchConfig with reboot settings and update classifications.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const patchDeployment = new gcp.osconfig.PatchDeployment("weekly-patch", {
        patchDeploymentId: "weekly-patches",
        instanceFilter: {
            all: true,
        },
        recurringSchedule: {
            timeZone: { id: "UTC" },
            weekly: { dayOfWeek: "SUNDAY" },
        },
        patchConfig: {
            rebootConfig: "DEFAULT",
        },
    });
    

    database-migration-service-not-publicly-accessible

    Severity: high · Enforcement: advisory

    Prevent public accessibility of Database Migration Service instances

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Configure connection profiles to use private connectivity (VPC peering or Private Service Connect) instead of public IP addresses. Ensure SSL is configured with CA certificates.

    Example - Secure Database Migration Service Connection Profiles:

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: MySQL connection profile with SSL and private IP
    const mysqlCompliant = new gcp.databasemigrationservice.ConnectionProfile("mysql-private", {
        location: "us-central1",
        connectionProfileId: "mysql-private-profile",
        displayName: "MySQL Private Connection",
        mysql: {
            host: "10.0.0.5",  // Compliant: private IP address
            port: 3306,
            username: "migration-user",
            password: sqldbUser.password,
            ssl: {  // Compliant: SSL configured
                clientKey: sqlClientCert.privateKey,
                clientCertificate: sqlClientCert.cert,
                caCertificate: sqlClientCert.serverCaCert,
                type: "SERVER_CLIENT"
            }
        }
    });
    
    // COMPLIANT: PostgreSQL with internal hostname and SSL
    const postgresCompliant = new gcp.databasemigrationservice.ConnectionProfile("postgres-private", {
        location: "us-central1",
        connectionProfileId: "postgres-private-profile",
        displayName: "PostgreSQL Private Connection",
        postgresql: {
            host: "db.internal.example.com",  // Compliant: internal hostname
            port: 5432,
            username: "migration-user",
            password: pgUser.password,
            ssl: {  // Compliant: SSL configured
                clientCertificate: pgClientCert.cert,
                caCertificate: pgClientCert.serverCaCert
            }
        }
    });
    
    // NON-COMPLIANT: MySQL connection without SSL
    const mysqlNoSsl = new gcp.databasemigrationservice.ConnectionProfile("mysql-no-ssl", {
        location: "us-central1",
        connectionProfileId: "mysql-insecure",
        mysql: {
            host: "203.0.113.10",  // Public IP
            port: 3306,
            username: "user",
            password: "password"
            // Missing ssl configuration
        }
    });
    
    // NON-COMPLIANT: PostgreSQL with public IP and no SSL
    const postgresPublic = new gcp.databasemigrationservice.ConnectionProfile("postgres-public", {
        location: "us-central1",
        connectionProfileId: "postgres-public",
        postgresql: {
            host: "198.51.100.20",  // Non-compliant: public IP
            port: 5432,
            username: "user",
            password: "password"
            // Missing ssl configuration
        }
    });
    

    dataproc-cluster-master-nodes-no-public-ip

    Severity: high · Enforcement: advisory

    Restrict public access for Dataproc clusters by ensuring master nodes use internal IP addresses only

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Set clusterConfig.gceClusterConfig.internalIpOnly: true for the Dataproc cluster. Enable Private Google Access on the VPC subnet. Use Cloud NAT or VPN/Interconnect for required external connectivity. Configure firewall rules for internal communication.

    import * as gcp from "@pulumi/gcp";
    
    const privateDataprocCluster = new gcp.dataproc.Cluster("private-dataproc", {
        name: "private-dataproc-cluster",
        region: "us-central1",
        clusterConfig: {
            gceClusterConfig: {
                // Disable external IPs for all nodes including master
                internalIpOnly: true,
                network: "projects/my-project/global/networks/my-vpc",
                subnetwork: "projects/my-project/regions/us-central1/subnetworks/my-subnet",
                zone: "us-central1-a",
                // Optional: Configure service account with necessary permissions
                serviceAccount: "dataproc-sa@my-project.iam.gserviceaccount.com",
                serviceAccountScopes: [
                    "https://www.googleapis.com/auth/cloud-platform",
                ],
            },
            masterConfig: {
                numInstances: 1,
                machineType: "n2-standard-4",
                diskConfig: {
                    bootDiskSizeGb: 100,
                    bootDiskType: "pd-standard",
                },
            },
            workerConfig: {
                numInstances: 2,
                machineType: "n2-standard-4",
            },
        },
    });
    

    dataproc-kerberos-enabled

    Severity: medium · Enforcement: advisory

    Ensure Dataproc Kerberos is enabled

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Configure clusterConfig.securityConfig.kerberosConfig with enableKerberos: true. Store Kerberos credentials in Secret Manager and provide URIs for rootPrincipalPasswordUri, keyPasswordUri, keystorePasswordUri, and truststorePasswordUri. Configure realm and KDC settings as needed.

    import * as gcp from "@pulumi/gcp";
    
    // First, create secrets in Secret Manager for Kerberos passwords
    const rootPassword = new gcp.secretmanager.Secret("kerberos-root-pwd", {
        secretId: "dataproc-kerberos-root-password",
        replication: { auto: {} },
    });
    
    const keystorePassword = new gcp.secretmanager.Secret("kerberos-keystore-pwd", {
        secretId: "dataproc-kerberos-keystore-password",
        replication: { auto: {} },
    });
    
    const truststorePassword = new gcp.secretmanager.Secret("kerberos-truststore-pwd", {
        secretId: "dataproc-kerberos-truststore-password",
        replication: { auto: {} },
    });
    
    const keyPassword = new gcp.secretmanager.Secret("kerberos-key-pwd", {
        secretId: "dataproc-kerberos-key-password",
        replication: { auto: {} },
    });
    
    // Create Dataproc cluster with Kerberos enabled
    const kerberosCluster = new gcp.dataproc.Cluster("kerberos-dataproc", {
        name: "secure-dataproc-cluster",
        region: "us-central1",
        clusterConfig: {
            securityConfig: {
                kerberosConfig: {
                    // Enable Kerberos authentication
                    enableKerberos: true,
                    // Reference secrets stored in Secret Manager
                    rootPrincipalPasswordUri: rootPassword.name.apply(name =>
                        `projects/my-project/secrets/${name}/versions/latest`
                    ),
                    keyPasswordUri: keyPassword.name.apply(name =>
                        `projects/my-project/secrets/${name}/versions/latest`
                    ),
                    keystorePasswordUri: keystorePassword.name.apply(name =>
                        `projects/my-project/secrets/${name}/versions/latest`
                    ),
                    truststorePasswordUri: truststorePassword.name.apply(name =>
                        `projects/my-project/secrets/${name}/versions/latest`
                    ),
                    // Optional: Configure realm and KDC
                    realm: "DATAPROC.EXAMPLE.COM",
                    // Optional: Cross-realm trust configuration
                    // crossRealmTrustRealm: "EXTERNAL.EXAMPLE.COM",
                    // crossRealmTrustKdc: "kdc.external.example.com",
                },
            },
            gceClusterConfig: {
                network: "projects/my-project/global/networks/my-vpc",
                subnetwork: "projects/my-project/regions/us-central1/subnetworks/my-subnet",
            },
        },
    });
    

    elasticsearch-encrypted-at-rest

    Severity: high · Enforcement: advisory

    Ensure Elasticsearch is encrypted at rest

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure bootDisk.kmsKeySelfLink for boot disks and diskEncryptionKeyRaw or kmsKeySelfLink for attached disks. Create Cloud KMS keys in the same region as the instance and grant the Compute Engine service account encrypt/decrypt permissions. All disks are encrypted by default with Google-managed keys; use CMEK for enhanced control.

    import * as gcp from "@pulumi/gcp";
    
    // Create KMS key ring and key for disk encryption
    const keyRing = new gcp.kms.KeyRing("elasticsearch-keyring", {
        name: "elasticsearch-disk-encryption",
        location: "us-central1",
    });
    
    const cryptoKey = new gcp.kms.CryptoKey("elasticsearch-key", {
        name: "elasticsearch-disk-key",
        keyRing: keyRing.id,
        rotationPeriod: "7776000s", // 90 days
    });
    
    // Grant Compute Engine service account permission to use the key
    const computeSA = gcp.compute.getDefaultServiceAccount({});
    const keyIAM = new gcp.kms.CryptoKeyIAMMember("compute-encrypter-decrypter", {
        cryptoKeyId: cryptoKey.id,
        role: "roles/cloudkms.cryptoKeyEncrypterDecrypter",
        member: computeSA.then(sa => `serviceAccount:${sa.email}`),
    });
    
    // Elasticsearch instance with CMEK-encrypted disks
    const encryptedElasticsearch = new gcp.compute.Instance("encrypted-elasticsearch", {
        name: "elasticsearch-encrypted",
        machineType: "n2-standard-8",
        zone: "us-central1-a",
        labels: {
            app: "elasticsearch",
        },
        // Boot disk with customer-managed encryption key
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
                size: 100,
            },
            kmsKeySelfLink: cryptoKey.id,
        },
        // Attached data disk with encryption
        attachedDisks: [{
            source: new gcp.compute.Disk("elasticsearch-data-disk", {
                name: "elasticsearch-data",
                zone: "us-central1-a",
                size: 500,
                type: "pd-ssd",
                diskEncryptionKey: {
                    kmsKeySelfLink: cryptoKey.id,
                },
            }).selfLink,
            kmsKeySelfLink: cryptoKey.id,
        }],
        networkInterfaces: [{
            network: "default",
        }],
    });
    
    // Alternative: Create separate persistent disk with encryption
    const elasticsearchDataDisk = new gcp.compute.Disk("elasticsearch-persistent-disk", {
        name: "elasticsearch-data-disk",
        zone: "us-central1-a",
        size: 1000, // 1TB for Elasticsearch data
        type: "pd-ssd",
        diskEncryptionKey: {
            kmsKeySelfLink: cryptoKey.id,
        },
    });
    

    elasticsearch-in-vpc-only

    Severity: high · Enforcement: advisory

    Ensure Elasticsearch is in VPC only

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove accessConfigs from networkInterfaces to disable external IP assignment for instances running Elasticsearch. Ensure instances are deployed within a VPC subnet. Access Elasticsearch through private networking or VPN/Interconnect only.

    import * as gcp from "@pulumi/gcp";
    
    // Elasticsearch instance with no external IP (VPC-only access)
    const elasticsearchInstance = new gcp.compute.Instance("elasticsearch-node", {
        name: "elasticsearch-node-1",
        machineType: "n2-standard-8",
        zone: "us-central1-a",
        labels: {
            environment: "production",
        },
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
                size: 100,
            },
        },
        networkInterfaces: [{
            network: "projects/my-project/global/networks/my-vpc",
            subnetwork: "projects/my-project/regions/us-central1/subnetworks/private-subnet",
            // IMPORTANT: No accessConfigs means no external IP
            // Instance is only accessible within VPC or via VPN/Interconnect
        }],
        // Optional: Add startup script to install Elasticsearch
        metadataStartupScript: `#!/bin/bash
            # Install and configure Elasticsearch
            wget -qO - https://artifacts.elastic.co/GPG-KEY-elasticsearch | apt-key add -
            echo "deb https://artifacts.elastic.co/packages/8.x/apt stable main" | tee /etc/apt/sources.list.d/elastic-8.x.list
            apt-get update && apt-get install -y elasticsearch
            systemctl enable elasticsearch
            systemctl start elasticsearch
        `,
        serviceAccount: {
            email: "elasticsearch-sa@my-project.iam.gserviceaccount.com",
            scopes: ["https://www.googleapis.com/auth/cloud-platform"],
        },
    });
    
    // Optional: Create firewall rule to allow Elasticsearch access from within VPC
    const elasticsearchFirewall = new gcp.compute.Firewall("elasticsearch-internal", {
        network: "projects/my-project/global/networks/my-vpc",
        allows: [{
            protocol: "tcp",
            ports: ["9200", "9300"],
        }],
        sourceRanges: ["10.0.0.0/8"], // Only allow from private IP ranges
        targetTags: ["elasticsearch"],
    });
    

    elasticsearch-logs-to-cloud-logging

    Severity: medium · Enforcement: advisory

    Ensure instances have proper Cloud Logging configuration

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Configure Compute Engine instances with the Ops Agent for log forwarding. Ensure service accounts have proper scopes for Cloud Logging API access. Install logging agents via startup scripts or custom images.

    import * as gcp from "@pulumi/gcp";
    
    // Service account with logging permissions
    const loggingSA = new gcp.serviceaccount.Account("logging-sa", {
        accountId: "instance-logger",
        displayName: "Instance Logging Service Account",
    });
    
    // Grant logging.logWriter role to service account
    const loggingRole = new gcp.projects.IAMMember("logging-role", {
        project: "my-project",
        role: "roles/logging.logWriter",
        member: loggingSA.email.apply(email => `serviceAccount:${email}`),
    });
    
    // Instance with proper logging configuration
    const instanceWithLogging = new gcp.compute.Instance("instance-with-logging", {
        name: "app-instance",
        machineType: "n2-standard-4",
        zone: "us-central1-a",
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
            },
        },
        networkInterfaces: [{
            network: "default",
            subnetwork: "default",
        }],
        serviceAccount: {
            email: loggingSA.email,
            // IMPORTANT: Include cloud-platform scope for logging APIs
            scopes: ["https://www.googleapis.com/auth/cloud-platform"],
        },
        // Install Ops Agent for log forwarding
        metadataStartupScript: `#!/bin/bash
            # Install Google Cloud Ops Agent
            curl -sSO https://dl.google.com/cloudagents/add-google-cloud-ops-agent-repo.sh
            bash add-google-cloud-ops-agent-repo.sh --also-install
    
            # Configure Ops Agent for application logs
            cat > /etc/google-cloud-ops-agent/config.yaml <<EOF
    logging:
      receivers:
        app_logs:
          type: files
          include_paths:
            - /var/log/app/*.log
            - /var/log/app/*.json
      processors:
        app_parser:
          type: parse_json
      service:
        pipelines:
          app_pipeline:
            receivers: [app_logs]
            processors: [app_parser]
    EOF
    
            # Restart Ops Agent to apply configuration
            systemctl restart google-cloud-ops-agent
        `,
        // Alternative: Use metadata for logging configuration
        metadata: {
            "google-logging-enabled": "true",
            "google-monitoring-enabled": "true",
        },
    });
    

    gke-cluster-endpoint-restrict-public-access

    Severity: high · Enforcement: advisory

    Restrict public access for GKE clusters by enabling private endpoint or configuring master authorized networks

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Enable private endpoint by setting privateClusterConfig.enablePrivateEndpoint: true, or configure masterAuthorizedNetworksConfig.cidrBlocks with specific IP ranges (avoid 0.0.0.0/0). Consider also enabling privateClusterConfig.enablePrivateNodes: true for node isolation.

    Pulumi TypeScript example:

    import * as gcp from "@pulumi/gcp";
    
    // Option 1: Private GKE cluster (recommended)
    const privateCluster = new gcp.container.Cluster("private-gke", {
        location: "us-central1",
        // Enable private cluster configuration
        privateClusterConfig: {
            enablePrivateEndpoint: true,  // Control plane only accessible via private IPs
            enablePrivateNodes: true,     // Nodes use private IPs only
            masterIpv4CidrBlock: "172.16.0.0/28",
        },
        // Required for private clusters
        ipAllocationPolicy: {},
        initialNodeCount: 1,
    });
    
    // Option 2: Public cluster with authorized networks
    const authorizedCluster = new gcp.container.Cluster("authorized-gke", {
        location: "us-central1",
        // Restrict access to specific IP ranges
        masterAuthorizedNetworksConfig: {
            cidrBlocks: [
                {
                    cidrBlock: "10.0.0.0/8",
                    displayName: "Corporate Network",
                },
                {
                    cidrBlock: "192.168.1.0/24",
                    displayName: "VPN Gateway",
                },
            ],
        },
        initialNodeCount: 1,
    });
    

    iam-no-custom-role-excessive-permissions

    Severity: high · Enforcement: advisory

    Prevent IAM custom roles with excessive permissions for enhanced data access control.

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Review and refine custom role permissions. Remove excessive administrative permissions like setIamPolicy, admin-level permissions, and owner permissions. Follow the principle of least privilege by granting only the minimum permissions required for the role’s intended purpose. Consider breaking up roles with too many permissions.

    Example - Create custom role with limited permissions:

    import * as gcp from "@pulumi/gcp";
    
    const role = new gcp.projects.IAMCustomRole("limited-role", {
        roleId: "limitedViewer",
        title: "Limited Viewer",
        permissions: [
            "compute.instances.get",
            "compute.instances.list",
        ],
    });
    

    iam-policy-no-statements-with-admin-access

    Severity: critical · Enforcement: advisory

    Prevent IAM policy bindings with administrative access permissions for secure configuration.

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Review and refine IAM policy bindings. Replace service-specific administrative roles like ‘roles/compute.admin’, ‘roles/storage.admin’, ‘roles/bigquery.admin’ with specific, granular roles. Use predefined roles with minimal permissions required for the task. Follow the principle of least privilege by granting only necessary permissions for operations, not administrative control over entire services.

    Example - Replace admin role with specific permissions:

    import * as gcp from "@pulumi/gcp";
    
    // Instead of roles/compute.admin, use specific roles
    const binding = new gcp.projects.IAMMember("compute-viewer", {
        project: "my-project",
        role: "roles/compute.viewer",
        member: "serviceAccount:sa@project.iam.gserviceaccount.com",
    });
    

    iam-policy-no-statements-with-full-access

    Severity: critical · Enforcement: advisory

    Prevent IAM policy bindings with full access permissions for enhanced data security.

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Review and refine IAM policy bindings. Replace broad roles like ‘roles/owner’ and ‘roles/editor’ with specific, granular roles. Use predefined roles with minimal permissions required. Follow the principle of least privilege by granting only necessary permissions.

    Example - Replace owner/editor with specific roles:

    import * as gcp from "@pulumi/gcp";
    
    // Instead of roles/editor, use specific roles
    const binding = new gcp.projects.IAMMember("storage-admin", {
        project: "my-project",
        role: "roles/storage.objectAdmin",
        member: "user:admin@example.com",
    });
    

    iam-policy-no-wildcard-permissions

    Severity: critical · Enforcement: advisory

    Prevent IAM custom roles with wildcard permissions to enforce least privilege access control.

    • 5.4 — Restrict administrator privileges to dedicated administrator accounts on enterprise assets. Conduct general computing activities, such as internet browsing, email, and productivity suite use, from the user’s primary, non-privileged account.
    Remediation

    Review and refine custom IAM roles. Remove wildcard () permissions and replace them with specific, granular permissions. Instead of using ‘compute.’ or ‘storage.*’, explicitly list the required permissions such as ‘compute.instances.get’, ‘compute.instances.list’, ‘storage.buckets.get’, etc. Follow the principle of least privilege by granting only the exact permissions needed for the role’s intended purpose.

    Example - Create custom role with specific permissions:

    import * as gcp from "@pulumi/gcp";
    
    const customRole = new gcp.organizations.IAMCustomRole("custom-role", {
        roleId: "computeViewer",
        orgId: "123456789",
        title: "Compute Viewer",
        permissions: [
            "compute.instances.get",
            "compute.instances.list",
        ],
    });
    

    iam-user-no-direct-policies-check

    Severity: high · Enforcement: advisory

    Prevent IAM users from having direct policy attachments for better access management.

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove direct IAM policy bindings from individual user and service accounts. Instead, create IAM groups and assign users to groups, or use predefined roles. For service accounts, grant permissions through workload identity federation or service account impersonation where possible. This improves manageability and follows the principle of least privilege.

    Example - Use groups instead of direct user bindings:

    import * as gcp from "@pulumi/gcp";
    
    // Grant permissions to a group instead of individual users
    const groupBinding = new gcp.projects.IAMMember("group-viewer", {
        project: "my-project",
        role: "roles/viewer",
        member: "group:dev-team@example.com",
    });
    

    kms-key-not-scheduled-for-deletion

    Severity: high · Enforcement: advisory

    Ensure Cloud KMS keys are not scheduled for deletion

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Cancel any pending key deletions by removing destroyScheduledDuration. Implement key rotation policies instead of destroying active encryption keys to maintain data access.

    Example - Create KMS key without scheduled deletion:

    import * as gcp from "@pulumi/gcp";
    
    const key = new gcp.kms.CryptoKey("my-key", {
        name: "crypto-key",
        keyRing: keyring.id,
        rotationPeriod: "7776000s",
        // Do not set destroyScheduledDuration
    });
    

    kms-key-rotation-enabled

    Severity: high · Enforcement: advisory

    Ensure Cloud KMS key rotation is enabled

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Enable automatic key rotation for Cloud KMS crypto keys by setting ‘rotationPeriod’. Key rotation limits the lifetime of encryption keys and reduces cryptographic risk.

    Example - Enable automatic key rotation:

    import * as gcp from "@pulumi/gcp";
    
    const key = new gcp.kms.CryptoKey("my-key", {
        name: "crypto-key",
        keyRing: keyring.id,
        rotationPeriod: "7776000s", // 90 days (configurable)
    });
    

    Common rotation periods:

    • 30 days: “2592000s”
    • 60 days: “5184000s”
    • 90 days: “7776000s” (recommended default)
    • 365 days: “31536000s” (maximum recommended)

    load-balancer-cloud-armor-enabled

    Severity: high · Enforcement: advisory

    Ensure Load Balancer has Cloud Armor enabled for network monitoring and defense

    • 13.1 — Centralize security event alerting across enterprise assets for log correlation and analysis. Best practice implementation requires the use of a SIEM, which includes vendor-defined event correlation alerts. A log analytics platform configured with security-relevant correlation alerts also satisfies this Safeguard.
    Remediation

    Create a Cloud Armor security policy using gcp.compute.SecurityPolicy and attach it to backend services via the securityPolicy property. Configure rules for DDoS protection, rate limiting, and geographic restrictions. Enable logging for security monitoring.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create Cloud Armor security policy
    const securityPolicy = new gcp.compute.SecurityPolicy("cloud-armor", {
        name: "my-security-policy",
        description: "Cloud Armor policy for DDoS protection and rate limiting",
        adaptiveProtectionConfig: {
            layer7DdosDefenseConfig: {
                enable: true
            }
        },
        rules: [
            {
                action: "rate_based_ban",
                priority: 1000,
                match: {
                    versionedExpr: "SRC_IPS_V1",
                    config: {
                        srcIpRanges: ["*"]
                    }
                },
                rateLimitOptions: {
                    conformAction: "allow",
                    exceedAction: "deny(429)",
                    rateLimitThreshold: {
                        count: 100,
                        intervalSec: 60
                    },
                    banDurationSec: 600
                },
                description: "Rate limit: 100 requests per minute"
            },
            {
                action: "deny(403)",
                priority: 2000,
                match: {
                    versionedExpr: "SRC_IPS_V1",
                    config: {
                        srcIpRanges: ["203.0.113.0/24"]  // Block malicious IP range
                    }
                },
                description: "Block known malicious IPs"
            },
            {
                action: "allow",
                priority: 2147483647,
                match: {
                    versionedExpr: "SRC_IPS_V1",
                    config: {
                        srcIpRanges: ["*"]
                    }
                },
                description: "Default allow rule"
            }
        ]
    });
    
    // Attach Cloud Armor to backend service
    const backendService = new gcp.compute.BackendService("backend", {
        name: "protected-backend",
        securityPolicy: securityPolicy.id,
        loadBalancingScheme: "EXTERNAL_MANAGED",
        healthChecks: healthCheck.id,
        backends: [{
            group: instanceGroup.id,
            balancingMode: "UTILIZATION"
        }],
        logConfig: {
            enable: true,
            sampleRate: 1.0
        }
    });
    

    load-balancer-cross-region-load-balancing-enabled

    Severity: medium · Enforcement: advisory

    Ensure Load Balancer cross-region load balancing is enabled

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Use Global Load Balancers with loadBalancingScheme: EXTERNAL_MANAGED. Configure backend services with multiple backends across different regions. Configure health checks for automatic failover. Set connectionDrainingTimeoutSec for graceful handling of backend failures.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create health check for automatic failover
    const healthCheck = new gcp.compute.HttpHealthCheck("health-check", {
        name: "multi-region-health-check",
        requestPath: "/health",
        checkIntervalSec: 5,
        timeoutSec: 3,
        healthyThreshold: 2,
        unhealthyThreshold: 2
    });
    
    // Create instance groups in multiple regions
    const usEastGroup = new gcp.compute.InstanceGroupManager("us-east", {
        name: "instance-group-us-east",
        baseInstanceName: "vm-us-east",
        zone: "us-east1-b",
        // ... instance template configuration
    });
    
    const usWestGroup = new gcp.compute.InstanceGroupManager("us-west", {
        name: "instance-group-us-west",
        baseInstanceName: "vm-us-west",
        zone: "us-west1-a",
        // ... instance template configuration
    });
    
    // Create global backend service with cross-region backends
    const backendService = new gcp.compute.BackendService("multi-region-backend", {
        name: "cross-region-backend",
        loadBalancingScheme: "EXTERNAL_MANAGED",
        healthChecks: healthCheck.id,
        connectionDrainingTimeoutSec: 300,
        backends: [
            {
                group: usEastGroup.instanceGroup,
                balancingMode: "UTILIZATION",
                capacityScaler: 1.0,
                maxUtilization: 0.8
            },
            {
                group: usWestGroup.instanceGroup,
                balancingMode: "UTILIZATION",
                capacityScaler: 1.0,
                maxUtilization: 0.8
            }
        ],
        logConfig: {
            enable: true,
            sampleRate: 1.0
        }
    });
    
    // Create URL map and forwarding rule
    const urlMap = new gcp.compute.URLMap("url-map", {
        name: "multi-region-url-map",
        defaultService: backendService.id
    });
    
    const httpsProxy = new gcp.compute.TargetHttpsProxy("https-proxy", {
        name: "https-proxy",
        urlMap: urlMap.id,
        sslCertificates: [sslCert.id]
    });
    
    const forwardingRule = new gcp.compute.GlobalForwardingRule("forwarding-rule", {
        name: "global-forwarding-rule",
        target: httpsProxy.id,
        portRange: "443",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL_MANAGED"
    });
    

    load-balancer-deletion-protection-enabled

    Severity: high · Enforcement: advisory

    Ensure Load Balancer has high availability configuration

    • 12.2 — Design and maintain a secure network architecture. A secure network architecture must address segmentation, least privilege, and availability, at a minimum. Example implementations may include documentation, policy, and design components.
    Remediation

    Configure load balancers for high availability:

    • Backend services should have multiple backend groups across availability zones
    • Health checks should be configured for automatic failover
    • Connection draining timeout should be set for graceful shutdowns
    • Consider using global load balancers for cross-region failover

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create health check for availability monitoring
    const healthCheck = new gcp.compute.HealthCheck("health-check", {
        name: "backend-health-check",
        checkIntervalSec: 10,
        timeoutSec: 5,
        healthyThreshold: 2,
        unhealthyThreshold: 3,
        httpHealthCheck: {
            port: 80,
            requestPath: "/health"
        }
    });
    
    // Create backend service with high availability configuration
    const backendService = new gcp.compute.BackendService("backend", {
        name: "ha-backend",
        healthChecks: healthCheck.id,
        connectionDrainingTimeoutSec: 30,  // Graceful connection draining
        loadBalancingScheme: "EXTERNAL",
        protocol: "HTTP",
        timeoutSec: 30,
        backends: [
            {
                group: instanceGroup1.id,
                balancingMode: "UTILIZATION",
                capacityScaler: 1.0,
                maxUtilization: 0.8
            },
            {
                group: instanceGroup2.id,  // Multiple backend groups for HA
                balancingMode: "UTILIZATION",
                capacityScaler: 1.0,
                maxUtilization: 0.8
            }
        ]
    });
    
    // Regional backend service with health checks
    const regionalBackend = new gcp.compute.RegionBackendService("regional-backend", {
        name: "regional-ha-backend",
        region: "us-central1",
        healthChecks: healthCheck.id,
        connectionDrainingTimeoutSec: 30,
        loadBalancingScheme: "INTERNAL",
        backends: [
            {
                group: regionalInstanceGroup1.id,
                balancingMode: "CONNECTION"
            },
            {
                group: regionalInstanceGroup2.id,
                balancingMode: "CONNECTION"
            }
        ]
    });
    

    load-balancer-http-to-https-redirection

    Severity: high · Enforcement: advisory

    Ensure Load Balancer HTTP to HTTPS redirection is configured

    • 3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).
    Remediation

    Configure URL map with HTTPS redirect by setting defaultUrlRedirect.httpsRedirect: true or configure path matchers with HTTPS redirects. Create separate HTTP and HTTPS target proxies, with HTTP proxy redirecting to HTTPS.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create URL map with HTTP to HTTPS redirect
    const redirectUrlMap = new gcp.compute.URLMap("redirect-map", {
        name: "http-redirect",
        defaultUrlRedirect: {
            httpsRedirect: true,
            redirectResponseCode: "MOVED_PERMANENTLY_DEFAULT"
        }
    });
    
    // Create HTTP target proxy for redirection
    const httpProxy = new gcp.compute.TargetHttpProxy("http-proxy", {
        name: "http-redirect-proxy",
        urlMap: redirectUrlMap.id
    });
    
    // Create HTTP forwarding rule (port 80)
    const httpForwardingRule = new gcp.compute.GlobalForwardingRule("http-rule", {
        name: "http-forwarding-rule",
        target: httpProxy.id,
        portRange: "80",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL"
    });
    
    // Create HTTPS URL map with backend service
    const httpsUrlMap = new gcp.compute.URLMap("https-map", {
        name: "https-url-map",
        defaultService: backendService.id
    });
    
    // Create HTTPS target proxy
    const httpsProxy = new gcp.compute.TargetHttpsProxy("https-proxy", {
        name: "https-proxy",
        urlMap: httpsUrlMap.id,
        sslCertificates: [sslCert.id]
    });
    
    // Create HTTPS forwarding rule (port 443)
    const httpsForwardingRule = new gcp.compute.GlobalForwardingRule("https-rule", {
        name: "https-forwarding-rule",
        target: httpsProxy.id,
        portRange: "443",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL"
    });
    

    load-balancer-logging-enabled

    Severity: high · Enforcement: advisory

    Enable Load Balancer logging for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Enable logging on backend services by configuring logConfig.enable: true and set logConfig.sampleRate (1.0 for full logging, lower values for sampling). Ensure logs are retained in Cloud Logging for security analysis and troubleshooting.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // For global backend service
    const backendService = new gcp.compute.BackendService("my-backend", {
        logConfig: {
            enable: true,
            sampleRate: 1.0  // 1.0 = 100% logging, 0.5 = 50% sampling
        },
        // ... other configuration
    });
    
    // For regional backend service
    const regionBackendService = new gcp.compute.RegionBackendService("my-region-backend", {
        logConfig: {
            enable: true,
            sampleRate: 1.0
        },
        // ... other configuration
    });
    

    load-balancer-managed-ssl-certificate

    Severity: high · Enforcement: advisory

    Ensure Load Balancer uses managed SSL certificates

    • 3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).
    Remediation

    Create managed SSL certificates using gcp.compute.ManagedSslCertificate with domains specified. Attach certificates to HTTPS target proxies via sslCertificates property. Google will automatically provision and renew certificates.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create Google-managed SSL certificate
    const managedCert = new gcp.compute.ManagedSslCertificate("managed-cert", {
        name: "auto-renew-cert",
        managed: {
            domains: [
                "example.com",
                "www.example.com",
                "api.example.com"
            ]
        }
    });
    
    // Create backend service
    const backendService = new gcp.compute.BackendService("backend", {
        name: "my-backend-service",
        healthChecks: healthCheck.id,
        backends: [{
            group: instanceGroup.id,
            balancingMode: "UTILIZATION"
        }]
    });
    
    // Create URL map
    const urlMap = new gcp.compute.URLMap("url-map", {
        name: "lb-url-map",
        defaultService: backendService.id
    });
    
    // Create HTTPS proxy with managed certificate
    const httpsProxy = new gcp.compute.TargetHttpsProxy("https-proxy", {
        name: "lb-https-proxy",
        urlMap: urlMap.id,
        sslCertificates: [managedCert.id]
    });
    
    // Create global forwarding rule
    const forwardingRule = new gcp.compute.GlobalForwardingRule("https-rule", {
        name: "lb-forwarding-rule",
        target: httpsProxy.id,
        portRange: "443",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL"
    });
    
    // Regional managed SSL certificate example
    const regionalHttpsProxy = new gcp.compute.RegionTargetHttpsProxy("regional-proxy", {
        name: "regional-https-proxy",
        region: "us-central1",
        urlMap: regionalUrlMap.id,
        sslCertificates: [managedCert.id]
    });
    

    load-balancer-tls-https-listeners-only

    Severity: high · Enforcement: advisory

    Ensure Load Balancer uses TLS/HTTPS listeners only

    • 3.10 — Encrypt sensitive data in transit. Example implementations can include: Transport Layer Security (TLS) and Open Secure Shell (OpenSSH).
    Remediation

    Configure forwarding rules to use port 443 for HTTPS traffic. Use TargetHttpsProxy instead of TargetHttpProxy. For HTTP listeners, ensure they redirect to HTTPS via URL map configuration. Remove standalone HTTP listeners without HTTPS redirection.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create SSL certificate (or use existing)
    const sslCert = new gcp.compute.ManagedSslCertificate("my-cert", {
        managed: {
            domains: ["example.com"]
        }
    });
    
    // Create HTTPS target proxy
    const httpsProxy = new gcp.compute.TargetHttpsProxy("https-proxy", {
        urlMap: urlMap.id,
        sslCertificates: [sslCert.id]
    });
    
    // Create forwarding rule on port 443
    const httpsForwardingRule = new gcp.compute.GlobalForwardingRule("https-rule", {
        target: httpsProxy.id,
        portRange: "443",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL"
    });
    
    // Optional: HTTP to HTTPS redirect
    const httpProxy = new gcp.compute.TargetHttpProxy("http-proxy", {
        urlMap: redirectUrlMap.id  // URL map configured for redirect
    });
    
    const httpForwardingRule = new gcp.compute.GlobalForwardingRule("http-rule", {
        target: httpProxy.id,
        portRange: "80",
        ipProtocol: "TCP",
        loadBalancingScheme: "EXTERNAL"
    });
    

    managed-instance-group-launch-template-public-ip-disabled

    Severity: high · Enforcement: advisory

    Ensure Managed Instance Group launch templates have public IP addresses disabled

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Disable public IP in instance template:

    new gcp.compute.InstanceTemplate("my-template", {
      networkInterfaces: [{
        network: "my-vpc",
        subnetwork: "my-subnet",
        accessConfigs: [], // Empty array disables public IP
      }],
      // ... other config
    });
    

    This prevents unintended public accessibility.

    managed-instance-group-load-balancer-healthcheck-required

    Severity: medium · Enforcement: advisory

    Ensure Managed Instance Groups have Load Balancer health check required

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Configure health checks for Managed Instance Groups by setting ‘autoHealingPolicies’ with a health check resource. This ensures that unhealthy instances are automatically replaced, improving availability and reliability.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const healthCheck = new gcp.compute.HealthCheck("http-health-check", {
        httpHealthCheck: { port: 80 },
    });
    
    const mig = new gcp.compute.InstanceGroupManager("my-mig", {
        baseInstanceName: "app",
        zone: "us-central1-a",
        autoHealingPolicies: {
            healthCheck: healthCheck.id,
            initialDelaySec: 300,
        },
        targetPools: [/* load balancer target pools */],
        versions: [{ instanceTemplate: /* template */ }],
    });
    

    no-unrestricted-route-to-internet-gateway

    Severity: high · Enforcement: advisory

    Ensure no unrestricted route to Internet Gateway for monitoring

    • 3.8 — Document data flows. Data flow documentation includes service provider data flows and should be based on the enterprise’s data management process. Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Route internet-bound traffic through Cloud NAT with logging enabled instead of direct internet gateway routes. Enable VPC Flow Logs for traffic monitoring. Use specific routes for known destinations. Configure firewall rules to control outbound traffic.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Bad: Direct route to default internet gateway (triggers violation)
    const directInternetRoute = new gcp.compute.Route("direct-internet-route", {
        network: network.id,
        destRange: "0.0.0.0/0",  // Default route
        nextHopGateway: "default-internet-gateway",  // Direct internet access
        priority: 1000,
    });
    
    // Good: Route through Cloud NAT for monitored egress
    const natRouter = new gcp.compute.Router("nat-router", {
        network: network.id,
        region: "us-central1",
    });
    
    const cloudNat = new gcp.compute.RouterNat("cloud-nat", {
        router: natRouter.name,
        region: natRouter.region,
        natIpAllocateOption: "AUTO_ONLY",
        sourceSubnetworkIpRangesToNat: "ALL_SUBNETWORKS_ALL_IP_RANGES",
        logConfig: {
            enable: true,
            filter: "ERRORS_ONLY",  // or "TRANSLATIONS_ONLY", "ALL"
        },
    });
    
    // Good: Specific route for known destination
    const specificRoute = new gcp.compute.Route("specific-route", {
        network: network.id,
        destRange: "10.128.0.0/20",  // Specific destination
        nextHopGateway: "default-internet-gateway",
        priority: 1000,
    });
    
    // Good: Route through internal load balancer for centralized egress
    const ilbRoute = new gcp.compute.Route("ilb-egress-route", {
        network: network.id,
        destRange: "0.0.0.0/0",
        nextHopIlb: ilb.selfLink,  // Centralized egress control
        priority: 1000,
    });
    

    Recommended patterns:

    1. Use Cloud NAT with logging enabled for internet-bound traffic
    2. Enable VPC Flow Logs on subnets for comprehensive monitoring
    3. Use specific routes (not 0.0.0.0/0) when possible
    4. Route through ILB for centralized egress control and monitoring

    persistent-disk-attached-encryption-enabled

    Severity: high · Enforcement: advisory

    Ensure Persistent Disks are encrypted when attached

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Configure persistent disks to use customer-managed encryption keys (CMEK) by setting the ‘diskEncryptionKey’ property with a KMS key. If CMEK is not required, Google-managed encryption is enabled by default.

    persistent-disk-encrypted

    Severity: high · Enforcement: advisory

    Ensure Persistent Disks are encrypted

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure diskEncryptionKey property with kmsKeySelfLink to use customer-managed encryption keys (CMEK) for enhanced security.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const disk = new gcp.compute.Disk("my-disk", {
        zone: "us-central1-a",
        type: "pd-standard",
        size: 10,
        diskEncryptionKey: {
            kmsKeySelfLink: "projects/my-project/locations/us-central1/keyRings/my-keyring/cryptoKeys/my-key",
        },
    });
    

    persistent-disk-in-backup-plan

    Severity: high · Enforcement: advisory

    Perform automated backups for Persistent Disks and maintain isolated recovery data

    • 11.2 — Perform automated backups of in-scope enterprise assets. Run backups weekly, or more frequently, based on the sensitivity of the data.
    • 11.4 — Establish and maintain an isolated instance of recovery data. Example implementations include version controlling backup destinations through offline, cloud, or off-site systems or services.
    Remediation

    Create a resource policy for automated disk snapshots and attach it to persistent disks using the resourcePolicies property with appropriate retention period.

    persistent-disk-snapshot-not-publicly-restorable

    Severity: high · Enforcement: advisory

    Restrict public access to Persistent Disk snapshots

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Remove IAM bindings granting allUsers or allAuthenticatedUsers permissions to snapshots. Use specific service accounts or groups for snapshot access.

    persistent-disk-unused

    Severity: low · Enforcement: advisory

    Ensure Persistent Disks are not unused

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Review unattached persistent disks and either attach them to instances or delete them if no longer needed. Unattached disks incur storage costs without providing value.

    project-part-of-organization

    Severity: medium · Enforcement: advisory

    Ensure project is part of GCP Organization

    • 4.1 — Establish and maintain a documented secure configuration process for enterprise assets (end-user devices, including portable and mobile, non-computing/IoT devices, and servers) and software (operating systems and applications). Review and update documentation annually, or when significant enterprise changes occur that could impact this Safeguard.
    Remediation

    Migrate the project to an organization using ‘gcloud projects move’. Projects should be part of an organization to enable centralized governance, billing, and security controls.

    Example - Project in Organization:

    import * as gcp from "@pulumi/gcp";
    
    // COMPLIANT: Project associated with an organization
    const projectCompliant = new gcp.organizations.Project("compliant-project", {
        name: "Production Project",
        projectId: "prod-project-123",
        orgId: "123456789012",  // Compliant: orgId specified
        labels: {
            environment: "production",
            team: "platform"
        },
        autoCreateNetwork: false
    });
    
    // NON-COMPLIANT: Project without organization association
    const projectNonCompliant = new gcp.organizations.Project("standalone-project", {
        name: "Standalone Project",
        projectId: "standalone-project-456"
        // Missing orgId - not associated with organization
    });
    
    // COMPLIANT: Project in organization with folder structure
    const projectInFolder = new gcp.organizations.Project("folder-project", {
        name: "Dev Project",
        projectId: "dev-project-789",
        orgId: "123456789012",  // Compliant: associated with organization
        folderId: "folders/987654321",  // Optional: further organization
        labels: {
            environment: "development"
        }
    });
    

    pubsub-encrypted-kms

    Severity: medium · Enforcement: advisory

    Ensure Pub/Sub is encrypted with Cloud KMS

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure kmsKeyName property when creating Pub/Sub topics to use customer-managed encryption keys (CMEK) for enhanced security and key management control.

    Example - Pub/Sub Topic with Customer-Managed Encryption:

    import * as gcp from "@pulumi/gcp";
    
    // Create KMS KeyRing and CryptoKey for encryption
    const keyRing = new gcp.kms.KeyRing("pubsub-keyring", {
        name: "pubsub-keyring",
        location: "us-central1"
    });
    
    const cryptoKey = new gcp.kms.CryptoKey("pubsub-key", {
        name: "pubsub-encryption-key",
        keyRing: keyRing.id,
        rotationPeriod: "7776000s"  // 90 days
    });
    
    // Grant Pub/Sub service account access to the key
    const binding = new gcp.kms.CryptoKeyIAMBinding("pubsub-key-binding", {
        cryptoKeyId: cryptoKey.id,
        role: "roles/cloudkms.cryptoKeyEncrypterDecrypter",
        members: ["serviceAccount:service-PROJECT_NUMBER@gcp-sa-pubsub.iam.gserviceaccount.com"]
    });
    
    // COMPLIANT: Topic with customer-managed encryption key
    const topicCompliant = new gcp.pubsub.Topic("encrypted-topic", {
        name: "secure-topic",
        kmsKeyName: cryptoKey.id,  // Compliant: CMEK configured
        messageRetentionDuration: "86400s"
    });
    
    // NON-COMPLIANT: Topic without CMEK (uses Google-managed keys)
    const topicNonCompliant = new gcp.pubsub.Topic("default-topic", {
        name: "insecure-topic"
        // Missing kmsKeyName - uses default Google-managed encryption
    });
    

    restricted-ssh

    Severity: high · Enforcement: advisory

    Restrict SSH access for monitoring compliance

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Restrict SSH access by updating firewall rule sourceRanges to specific trusted IP CIDR blocks instead of 0.0.0.0/0. Use Identity-Aware Proxy (IAP) with source range 35.235.240.0/20 for secure SSH access. Implement bastion hosts or VPN for remote access.

    Example Pulumi TypeScript:

    import * as gcp from "@pulumi/gcp";
    
    const sshFirewall = new gcp.compute.Firewall("restricted-ssh", {
        network: vpc.id,
        direction: "INGRESS",
        sourceRanges: [
            "10.0.0.0/8",        // Internal network
            "35.235.240.0/20",   // IAP for SSH
        ],
        allows: [{
            protocol: "tcp",
            ports: ["22"],
        }],
    });
    

    secret-manager-using-cmek

    Severity: high · Enforcement: advisory

    Ensure Secret Manager uses customer-managed encryption keys

    • 3.11 — Encrypt sensitive data at rest on servers, applications, and databases. Storage-layer encryption, also known as server-side encryption, meets the minimum requirement of this Safeguard. Additional encryption methods may include application-layer encryption, also known as client-side encryption, where access to the data storage device(s) does not permit access to the plain-text data.
    Remediation

    Configure Secret Manager secrets with customer-managed encryption keys (CMEK). Set the replication property with customerManagedEncryption and kmsKeyName for automatic replication, or configure customerManagedEncryption on each replica for user-managed replication. This provides enhanced security and key management control over sensitive secrets.

    import * as gcp from "@pulumi/gcp";
    
    // Create KMS keys for encryption
    const keyring = new gcp.kms.KeyRing("secret-keyring", {
        name: "secret-keyring",
        location: "us-central1"
    });
    
    const kmsKey = new gcp.kms.CryptoKey("secret-key", {
        name: "secret-key",
        keyRing: keyring.id
    });
    
    // COMPLIANT: Secret with auto replication and CMEK
    const secretAutoCompliant = new gcp.secretmanager.Secret("auto-cmek-secret", {
        secretId: "api-key",
        replication: {
            auto: {
                customerManagedEncryption: {
                    kmsKeyName: kmsKey.id  // Compliant: CMEK configured
                }
            }
        }
    });
    
    // COMPLIANT: Secret with user-managed replication and CMEK per region
    const secretUserManagedCompliant = new gcp.secretmanager.Secret("user-managed-cmek", {
        secretId: "database-password",
        replication: {
            userManaged: {
                replicas: [
                    {
                        location: "us-central1",
                        customerManagedEncryption: {
                            kmsKeyName: kmsKey.id  // Compliant: CMEK per replica
                        }
                    },
                    {
                        location: "us-east1",
                        customerManagedEncryption: {
                            kmsKeyName: kmsKeyEast.id
                        }
                    }
                ]
            }
        }
    });
    
    // NON-COMPLIANT: Secret without CMEK (uses Google-managed keys)
    const secretNonCompliant = new gcp.secretmanager.Secret("no-cmek-secret", {
        secretId: "api-token",
        replication: {
            auto: {}  // Missing customerManagedEncryption
        }
    });
    
    // NON-COMPLIANT: User-managed replication without CMEK
    const secretUserManagedNonCompliant = new gcp.secretmanager.Secret("no-cmek-user-managed", {
        secretId: "credentials",
        replication: {
            userManaged: {
                replicas: [
                    {
                        location: "us-central1"
                        // Missing customerManagedEncryption
                    }
                ]
            }
        }
    });
    

    security-command-center-enabled

    Severity: high · Enforcement: advisory

    Ensure Security Command Center is enabled

    • 1.2 — Ensure that a process exists to address unauthorized assets on a weekly basis. The enterprise may choose to remove the asset from the network, deny the asset from connecting remotely to the network, or quarantine the asset.
    Remediation

    Enable the Security Command Center API service (securitycenter.googleapis.com) in your project or organization. Configure notification channels using gcp.securitycenter.NotificationConfig resources to receive security alerts. Consider enabling Security Command Center’s built-in detectors and custom modules for comprehensive security monitoring.

    Example - Enable Security Command Center:

    import * as gcp from "@pulumi/gcp";
    
    const sccService = new gcp.projects.Service("scc-api", {
        service: "securitycenter.googleapis.com",
    });
    
    const notifConfig = new gcp.securitycenter.NotificationConfig("scc-notify", {
        configId: "scc-notifications",
        organization: "123456789",
        pubsubTopic: topic.id,
        streamingConfig: { filter: 'state = "ACTIVE"' },
    });
    

    subnet-auto-assign-external-ip-disabled

    Severity: medium · Enforcement: advisory

    Ensure Private Google Access is enabled on VPC subnets

    • 3.3 — Configure data access control lists based on a user’s need to know. Apply data access control lists, also known as access permissions, to local and remote file systems, databases, and applications.
    Remediation

    Enable Private Google Access on subnets by setting privateIpGoogleAccess: true. This allows instances without external IP addresses to access Google APIs and services. Use Cloud NAT for internet access without public IPs.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Enable Private Google Access on subnet
    const subnet = new gcp.compute.Subnetwork("my-subnet", {
        ipCidrRange: "10.0.0.0/24",
        network: network.id,
        region: "us-central1",
        privateIpGoogleAccess: true  // Allow instances without external IPs to access Google APIs
    });
    
    // Optional: Create Cloud NAT for outbound internet access
    const router = new gcp.compute.Router("my-router", {
        network: network.id,
        region: "us-central1"
    });
    
    const nat = new gcp.compute.RouterNat("my-nat", {
        router: router.name,
        region: router.region,
        natIpAllocateOption: "AUTO_ONLY",
        sourceSubnetworkIpRangesToNat: "ALL_SUBNETWORKS_ALL_IP_RANGES"
    });
    

    vpc-external-ip-associated

    Severity: medium · Enforcement: advisory

    Ensure VPC external IP addresses are associated

    • 1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.
    Remediation

    Review reserved external IP addresses and either associate them with resources (instances, forwarding rules) or release unused addresses. Use gcloud compute addresses list to identify unassociated addresses and gcloud compute addresses delete to remove them.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Create a static external IP address
    const staticIp = new gcp.compute.Address("my-static-ip", {
        name: "my-external-ip",
        region: "us-central1",
        addressType: "EXTERNAL",
    });
    
    // Associate it with a compute instance
    const instance = new gcp.compute.Instance("my-instance", {
        machineType: "f1-micro",
        zone: "us-central1-a",
        bootDisk: {
            initializeParams: {
                image: "debian-cloud/debian-11",
            },
        },
        networkInterfaces: [{
            network: "default",
            accessConfigs: [{
                natIp: staticIp.address,  // Associate the static IP
            }],
        }],
    });
    
    // Or associate with a forwarding rule
    const forwardingRule = new gcp.compute.ForwardingRule("my-forwarding-rule", {
        ipAddress: staticIp.address,  // Associate the static IP
        target: targetPool.selfLink,
        portRange: "80",
        region: "us-central1",
    });
    

    vpc-firewall-rule-associated-to-network

    Severity: high · Enforcement: mandatory

    Ensure VPC firewall rules are associated to networks

    • 1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.
    Remediation

    Associate firewall rules with VPC networks by specifying the network property with a valid VPC network reference (name, partial URL, or full URL). Ensure the network exists before creating the firewall rule.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    const firewallRule = new gcp.compute.Firewall("my-firewall-rule", {
        network: network.id,  // Reference to VPC network
        allows: [{
            protocol: "tcp",
            ports: ["80", "443"],
        }],
        sourceRanges: ["0.0.0.0/0"],
        targetTags: ["web-server"],
    });
    

    The network property can be specified as:

    • A resource reference: network.id or network.selfLink
    • A full URL: https://www.googleapis.com/compute/v1/projects/PROJECT/global/networks/NETWORK
    • A partial URL: projects/PROJECT/global/networks/NETWORK
    • A short name: network-name

    vpc-firewall-rule-restrict-ingress-ssh-all

    Severity: high · Enforcement: advisory

    Restrict SSH access from all IPs in firewall rules

    • 4.6 — Securely manage enterprise assets and software. Example implementations include managing configuration through version-controlled Infrastructure-as-Code (IaC) and accessing administrative interfaces over secure network protocols, such as Secure Shell (SSH) and Hypertext Transfer Protocol Secure (HTTPS). Do not use insecure management protocols, such as Telnet (Teletype Network) and HTTP, unless operationally essential.
    Remediation

    Restrict SSH access by updating firewall rules to allow SSH (port 22) only from specific trusted IP ranges. Remove rules that allow SSH from 0.0.0.0/0 or ::/0. Use Identity-Aware Proxy (IAP) or bastion hosts for secure SSH access.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Secure SSH rule - restrict to specific IPs
    const sshFirewall = new gcp.compute.Firewall("ssh-restricted", {
        network: "default",
        direction: "INGRESS",
        allows: [{
            protocol: "tcp",
            ports: ["22"]
        }],
        sourceRanges: [
            "203.0.113.0/24",  // Replace with your trusted IP ranges
            "198.51.100.0/24"
        ],
        targetTags: ["ssh-enabled"]
    });
    
    // Alternative: Use IAP for SSH (recommended)
    const iapSshFirewall = new gcp.compute.Firewall("ssh-iap", {
        network: "default",
        direction: "INGRESS",
        allows: [{
            protocol: "tcp",
            ports: ["22"]
        }],
        sourceRanges: ["35.235.240.0/20"],  // IAP IP range
        description: "Allow SSH via Identity-Aware Proxy"
    });
    

    vpc-firewall-rule-unused

    Severity: low · Enforcement: advisory

    Ensure VPC firewall rules are not unused

    • 1.1 — Establish and maintain an accurate, detailed, and up-to-date inventory of all enterprise assets with the potential to store or process data, to include: end-user devices (including portable and mobile), network devices, non-computing/IoT devices, and servers. Ensure the inventory records the network address (if static), hardware address, machine name, enterprise asset owner, department for each asset, and whether the asset has been approved to connect to the network. For mobile end-user devices, MDM type tools can support this process, where appropriate. This inventory includes assets connected to the infrastructure physically, virtually, remotely, and those within cloud environments. Additionally, it includes assets that are regularly connected to the enterprise’s network infrastructure, even if they are not under control of the enterprise. Review and update the inventory of all enterprise assets bi-annually, or more frequently.
    Remediation

    Review and remove disabled or unused firewall rules. Enable necessary rules or delete obsolete ones. Remove rules with test/temp/deprecated tags. Use gcloud compute firewall-rules list --filter='disabled=true' to identify disabled rules.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    // Good: Active firewall rule with clear purpose
    const activeFirewall = new gcp.compute.Firewall("web-allow-https", {
        network: network.id,
        disabled: false,  // Ensure rule is enabled
        allows: [{
            protocol: "tcp",
            ports: ["443"],
        }],
        sourceRanges: ["0.0.0.0/0"],
        targetTags: ["web-server"],
    });
    
    // Bad: Disabled firewall rule (should be removed or enabled)
    const disabledFirewall = new gcp.compute.Firewall("old-rule", {
        network: network.id,
        disabled: true,  // This will trigger a violation
        allows: [{
            protocol: "tcp",
            ports: ["8080"],
        }],
    });
    
    // Bad: Firewall rule with suspicious tags indicating it may be unused
    const testFirewall = new gcp.compute.Firewall("test-rule", {
        network: network.id,
        allows: [{
            protocol: "tcp",
            ports: ["22"],
        }],
        targetTags: ["test", "deprecated"],  // Will trigger a violation
    });
    
    // Use labels to track firewall rule usage and purpose
    const trackedFirewall = new gcp.compute.Firewall("tracked-rule", {
        network: network.id,
        allows: [{
            protocol: "tcp",
            ports: ["443"],
        }],
        sourceRanges: ["10.0.0.0/8"],
        targetTags: ["internal-api"],
    });
    

    To track firewall rule usage:

    1. Avoid using disabled: true - remove unused rules instead
    2. Don’t use tags like “test”, “temp”, “deprecated” in production
    3. Use descriptive names and regular audits to identify unused rules
    4. Review rules with high priority values (low precedence) that may never match

    vpc-flow-logs-enabled

    Severity: high · Enforcement: advisory

    Ensure VPC subnets have Flow Logs enabled for audit and monitoring

    • 8.2 — Collect audit logs. Ensure that logging, per the enterprise’s audit log management process, has been enabled across enterprise assets.
    Remediation

    Enable VPC Flow Logs on subnets by configuring logConfig with aggregationInterval (e.g., INTERVAL_5_SEC) and flowSampling (0.5 for 50% or 1.0 for 100%). Ensure flowSampling is greater than 0 to collect traffic logs.

    Pulumi TypeScript Example:

    import * as gcp from "@pulumi/gcp";
    
    const subnet = new gcp.compute.Subnetwork("my-subnet", {
        ipCidrRange: "10.0.0.0/24",
        network: network.id,
        region: "us-central1",
        logConfig: {
            aggregationInterval: "INTERVAL_5_SEC",  // or INTERVAL_30_SEC, INTERVAL_1_MIN, etc.
            flowSampling: 1.0,  // 1.0 = 100% sampling, 0.5 = 50%
            metadata: "INCLUDE_ALL_METADATA"
        },
        privateIpGoogleAccess: true
    });
    

    Valid aggregationInterval values: INTERVAL_5_SEC, INTERVAL_30_SEC, INTERVAL_1_MIN, INTERVAL_5_MIN, INTERVAL_10_MIN, INTERVAL_15_MIN

      The infrastructure as code platform for any cloud.