GCP Security Hardening

Enterprise-grade security hardening for MCP Server on GKE with 67 security controls and defense-in-depth

Overview

Implement defense-in-depth security for MCP Server on GKE with 67 security controls achieving CIS GKE Benchmark compliance and SOC 2 readiness.

Security Controls

67 controls across 7 layers

Compliance

CIS, SOC 2, HIPAA-ready

Zero Trust

Network policies, mTLS, Workload Identity

Automated Scanning

Daily compliance checks

---

7-Layer Security Architecture

1. Layer 1: Infrastructure

   • VPC isolation
   • IAM boundaries
   • GKE Security Posture Dashboard
   • Audit logging
2. Layer 2: Compute (Shielded Nodes)

   • Secure Boot (verify boot integrity)
   • vTPM (hardware-based key storage)
   • Integrity monitoring (detect tampering)

   Automatic in GKE Autopilot
3. Layer 3: Network

   • Private nodes (no public IPs)
   • VPC-native networking
   • Network policies (zero-trust)
   • Cloud Armor (DDoS protection)
4. Layer 4: Data

   • Encryption at rest (Google-managed or CMEK)
   • Encryption in transit (TLS required)
   • Secret Manager (no secrets in Git)
5. Layer 5: Identity

   • Workload Identity (no SA keys)
   • IAM least privilege
   • Master authorized networks
6. Layer 6: Application

   • Binary Authorization (image signing)
   • Container vulnerability scanning
   • Pod Security Standards (restricted)
7. Layer 7: Compliance

   • Audit logging
   • Policy enforcement (OPA/Gatekeeper)
   • Automated compliance scanning

---

Quick Security Setup

1. Enable Binary Authorization

1. Run Setup Script

   ./deployments/security/binary-authorization/setup-binary-auth.sh PROJECT_ID production
   

2. Sign Images

   ./deployments/security/binary-authorization/sign-image.sh \
     PROJECT_ID production IMAGE_URL
   

3. Enable in Cluster

   # terraform.tfvars
   enable_binary_authorization = true
   

   terraform apply
   

2. Configure Private Cluster

Private Nodes Only

enable_private_nodes = true     # ✅ Already configured
enable_private_endpoint = false # Public control plane access

Use case: Production with external access needed

3. Restrict Control Plane Access

enable_master_authorized_networks = true
master_authorized_networks_cidrs = [
  {
    cidr_block   = "10.0.0.0/8"
    display_name = "Internal VPC"
  },
  {
    cidr_block   = "203.0.113.0/24"
    display_name = "Office Network"
  }
]

4. Enable Security Posture Dashboard

enable_security_posture             = true
security_posture_mode               = "ENTERPRISE"
security_posture_vulnerability_mode = "VULNERABILITY_ENTERPRISE"

View Dashboard:

## Console: Kubernetes → Security
## https://console.cloud.google.com/kubernetes/security?project=PROJECT_ID

---

Security Checklist

Infrastructure ✅

• GKE Security Posture enabled
• Audit logging enabled
• Shielded nodes (automatic in Autopilot)
• Workload Identity enabled
• VPC Service Controls (optional, high-security)

Network ✅

• Private nodes (no public IPs)
• VPC-native networking
• Network policies enforced
• Cloud NAT for egress
• Private endpoint (optional)
• Cloud Armor (enable via variable)

Application ✅

• Binary Authorization ready
• Container scanning (Trivy in CI/CD)
• Pod security standards (restricted)
• Non-root containers
• Read-only root filesystem
• Capabilities dropped (ALL)

Data ✅

• Encryption at rest (Google-managed)
• CMEK (optional, compliance)
• Encryption in transit (TLS)
• Secrets in Secret Manager
• No secrets in Git/ConfigMaps
• No service account keys

Identity ✅

• Workload Identity (no SA keys)
• IAM least privilege
• Per-workload service accounts
• RBAC policies
• SA key creation disabled (org policy)

---

Security Testing

Automated Scans

Run daily compliance scans:

## Trigger manually
gh workflow run gcp-compliance-scan.yaml

## View results
gh run list --workflow=gcp-compliance-scan.yaml

Scans include:

• Terraform security (Trivy, tfsec, Checkov)
• Kubernetes manifests (Trivy, kube-score)
• CIS GKE Benchmark (kube-bench)
• Secret scanning (Gitleaks, TruffleHog)

Manual Security Review

Check for Privileged Containers

kubectl get pods -A -o json \
  | jq '[.items[] | select(.spec.containers[].securityContext.privileged==true)] | length'

Should return: 0

Verify Network Policies

kubectl get networkpolicies -n production-mcp-server-langgraph

At least 2 policies should exist

Audit Workload Identity

kubectl get sa -A -o json \
  | jq '.items[] | select(.metadata.annotations["iam.gke.io/gcp-service-account"]) | {namespace: .metadata.namespace, name: .metadata.name, gcp_sa: .metadata.annotations["iam.gke.io/gcp-service-account"]}'

Review IAM Bindings

gcloud projects get-iam-policy PROJECT_ID \
  --flatten="bindings[].members" \
  --filter="bindings.members:serviceAccount"

---

Compliance Frameworks

CIS GKE Benchmark

GKE Autopilot is pre-configured to meet CIS Benchmark requirements.

Key controls (auto-compliant):

• ✅ 4.1.1 Workload Identity enabled
• ✅ 4.2.1 Network Policy enabled
• ✅ 4.3.1 Private cluster enabled
• ✅ 4.5.1 Audit Logging enabled

Validate with kube-bench:

kubectl apply -f https://raw.githubusercontent.com/aquasecurity/kube-bench/main/job-gke.yaml
kubectl logs job/kube-bench

SOC 2 Type II

Required controls (all implemented):

• ✅ Access control (Workload Identity + RBAC)
• ✅ Encryption (at rest + in transit)
• ✅ Audit logging (all access logged)
• ✅ Change management (Terraform + GitOps)
• ✅ Monitoring & alerting (Cloud Operations)
• ✅ Incident response (runbooks documented)

HIPAA Compliance

Requirements:

• ✅ Encryption at rest (enable CMEK recommended)
• ✅ Encryption in transit (TLS required)
• ✅ Audit logs (configure 6+ year retention)
• ✅ Access controls (IAM + RBAC)
• ✅ Business Associate Agreement (sign with Google)

Additional steps:

## Enable CMEK
kms_key_name = "projects/PROJECT/locations/REGION/keyRings/RING/cryptoKeys/KEY"

## Extended audit log retention
## Configure via Cloud Console: Logging → Log Router → Sinks

---

Security Monitoring

Critical Metrics

Unauthorized Access

gcloud logging read \
  'protoPayload.status.code!=0 AND protoPayload.authenticationInfo.principalEmail!=""' \
  --limit=50

---

Incident Response

Security Incident Runbook

1. Detection

   • Security alert fires
   • Unusual activity in audit logs
   • Binary Authorization denial surge
2. Containment

   # Isolate affected pods
   kubectl label pod SUSPICIOUS_POD quarantine=true -n production-mcp-server-langgraph
   
   # Network policy to isolate
   kubectl apply -f - <<EOF
   apiVersion: networking.k8s.io/v1
   kind: NetworkPolicy
   metadata:
     name: isolate-suspicious-pod
     namespace: production-mcp-server-langgraph
   spec:
     podSelector:
       matchLabels:
         quarantine: "true"
     policyTypes:
     - Ingress
     - Egress
   EOF
   

3. Investigation

   # Export logs
   kubectl logs POD_NAME -n production-mcp-server-langgraph --all-containers > incident-logs.txt
   
   # Audit trail
   gcloud logging read \
     'resource.type="k8s_cluster" AND protoPayload.resourceName=~"pods/SUSPICIOUS_POD"' \
     --format=json > audit-trail.json
   

4. Remediation

   • Delete compromised pods
   • Rotate all secrets
   • Review and tighten IAM policies
   • Deploy patched version

---

Related Documentation

Binary Authorization

Image signing and policy enforcement

Operations Runbooks

Security operations and incident response

Compliance Scanning

Automated security and compliance scanning

Security Best Practices

Additional security hardening best practices