Skip to main content

Compliant

A Compliant system can demonstrate its controls and produce evidence on demand, without a two-week scramble for screenshots. Every edition of the Baseline evaluates its configuration continuously against the frameworks its code implements, keeps the audit trail where the account, subscription or project it describes cannot alter it, and gives auditors read access that cannot reach data. The code provides controls and evidence sources; it is not a certification or an attestation, which remain your auditor's decision.

How each edition delivers it​

Each edition evaluates with its cloud's own compliance service and names the frameworks it covers in code:

AspectAWS Enterprise BaselineAzure Enterprise BaselineGCP Enterprise Baseline
Evaluation serviceAWS Config conformance packs, Security Hub standardsAzure Policy at mg-acme, Defender for Cloud regulatory complianceOrganization policies, Security Command Center, log-based detections
On by defaultBuiltForProd-soc2-baseline conformance packCustom BuiltForProd-soc2-baseline initiative (Audit), Microsoft cloud security benchmark13 log-based detections, eight of them from the CIS Google Cloud Foundations Benchmark
Available as switches28 further packs, four Security Hub standardsNine more built-in initiatives, assigned in DoNotEnforceNone: no other framework is mapped in code
Excluded from scopeThe sandbox accountThe sandbox subscription, through recorded waiver exemptionsThe sandbox project, through a Security Command Center mute rule

AWS Enterprise Baseline​

AWS Config records every supported resource type in every account and region. The management account deploys the BuiltForProd-soc2-baseline conformance pack (encrypted volumes, multi-region CloudTrail, S3 SSL-only, MFA for the console, no root access keys, VPC Flow Logs and similar rules) to every account except sandbox, aggregated in the audit account; 28 further packs are switches. Security Hub in the security account aggregates findings across regions, with CIS AWS Foundations Benchmark v5.0.0, AWS Foundational Security Best Practices, PCI DSS v4.0.1 and the Resource Tagging standard as keys in security_hub_standards. The offerings note lists out-of-the-box readiness for SOC 2, HIPAA, PCI DSS 4.0, CMMC, NIST CSF and the CIS AWS Foundations Benchmark; the compliance guide delivered with the repositories maps each framework to what to enable:

FrameworkSecurity Hub standardsConformance packs
SOC 2cis_v5, fsbpsoc2-baseline (on by default), nist-csf, encryption-keys
CIS v5cis_v5cis-controls-v8-ig1, -ig2, -ig3
PCI DSS v4.0.1pci_dss_v4pci-dss-v4-incl-global
HIPAA and HITECHfsbphipaa-security
NIST SP 800-53 Rev 5fsbpnist-csf, encryption-keys
NIST SP 800-171 Rev 2fsbpnist-csf, cis-controls-v8-ig2
FedRAMP Moderatefsbp, pci_dss_v4nist-csf, encryption-keys
CMMC 2.0fsbpcmmc-2-level-1, cmmc-2-level-2
AWS Well-Architectedfsbpwa-security-pillar, wa-reliability-pillar

Each standard adds checks and therefore Security Hub cost, which is why they are off until a requirement exists. Evidence is produced read-only from the auditor profiles: the Security Hub score per standard, the conformance pack scores, Config resource history, and a quarterly snapshot of findings exported as JSON. External auditors receive a dedicated Identity Center group with ReadOnlyAccess to the security, audit and management accounts, never to workload accounts.

Azure Enterprise Baseline​

Azure Policy evaluates every subscription continuously from mg-acme. The custom initiative BuiltForProd-soc2-baseline, assigned as acme-soc2-base, covers 13 controls of the SOC 2 criteria CC6.1, CC6.6, CC6.7 and CC7.1 in Audit mode; every control's effect is a parameter, so one can be switched to Deny or Disabled per deployment. Two of its controls are custom definitions: no role assignments to users, and a customer-managed key on the audit storage (acme-audit-storage-cmk). The built-in framework initiatives are assigned in DoNotEnforce mode: their Audit and AuditIfNotExists policies evaluate, and their DeployIfNotExists and Modify policies report without acting. All are free to evaluate, and each is a key in compliance_initiatives in security.hcl:

FrameworkKeyDefault
BuiltForProd SOC 2 baseline (custom)soc2_baselineOn
Microsoft cloud security benchmarkmcsbOn
CIS Azure Foundations v3.0.0cis_azure_v3Off
NIST SP 800-53 Rev. 5nist_800_53_r5Off
NIST SP 800-171 Rev. 2nist_800_171_r2Off
PCI DSS v4pci_dss_v4Off
HITRUST/HIPAAhipaa_hitrustOff
SOC 2 Type 2 (built-in)soc2_type2Off
ISO 27001:2013iso_27001Off
CMMC 2.0 Level 2cmmc_l2Off
FedRAMP Moderatefedramp_moderateOff

acme-plat-sandbox is waived from every compliance assignment by an exemption of category Waiver with its reason in the description, so an auditor sees why it is out of scope. Evidence comes from the Policy compliance view per assignment, Defender for Cloud regulatory compliance and secure score exported to the security workspace, Kusto queries over the audit workspace, and the Git history of every change.

GCP Enterprise Baseline​

The GCP edition implements one framework control by control: the CIS Google Cloud Foundations Benchmark. Its logging and monitoring recommendations 2.4 to 2.11 are eight log-based metrics with alert policies over the central audit bucket, and DNS query logging (2.12) is on for every VPC. Five more detections are the platform's own: organization policy changes, service account key creation, break-glass sign-ins, HIGH and CRITICAL Security Command Center findings, and sink changes. The code maps no other framework.

ControlWhat it finds or preventsWhere
19 organization policy constraintsNon-compliant resources are refused at creation, project Owners includedOrganization node
Security Health Analytics (SCC Standard)Misconfigured resources: public buckets, open firewall rules, risky IAM grantsOrganization; scc_tier in security.hcl
CIS and platform log detectionsOwner, audit config, custom role, firewall, route, network, bucket IAM, policy and sink changescis-alerts in acme-core-audit
Custom label constraintsResources without the platform labels, in dry run until enforce_label_policyOrganization node

Keys rotate every 365 days; the module comment notes that CIS Google Cloud Foundations 1.10 asks for 90, a one-value change per key. Evidence comes from Security Command Center findings at the organization, with acme-plat-sandbox muted, from SQL queries on the central bucket through Log Analytics, from the alert history in acme-core-security, and from the Git history of every change.

Evidence that cannot be altered​

EvidenceAWS Enterprise BaselineAzure Enterprise BaselineGCP Enterprise Baseline
Audit trailOne organization CloudTrail with log file validation and KMS encryptionActivity Log of every subscription and Microsoft Entra ID logs, in the audit workspace (365 days)Admin Activity and Data Access audit logs, through two aggregated organization sinks
Where it is keptAudit bucket in the log-archive account: versioned, dedicated key, TLS-onlyAudit storage in acme-core-audit: write-once for 365 days, customer-managed key, cool and archive tiersacme-audit log bucket and the acme-usw1-audit-logs archive in acme-core-audit, with a customer-managed key
Protectionacme-audit-protection SCP on CloudTrail and ConfigDenyAction on diagnostic settings, the audit workspace and the audit storage; CanNotDelete locksIAM deny policy acme-audit-protection on sinks and audit buckets; prevent_destroy
Permanent lockCurrent objects are never deletedlock_immutability_policy locks the 365-day policy for goodlock_log_bucket_retention, lock_archive_bucket_retention lock the 365-day retention for good
Change historyPull requests, plans, run reportsPull requests, plans, run reportsPull requests, plans, run reports

The permanent locks are off by default: a locked policy cannot be shortened or removed, so it is turned on once the retention period is agreed with your auditors. The Observable page covers how the audit trail is collected.

Access governance​

ConcernAWS Enterprise BaselineAzure Enterprise BaselineGCP Enterprise Baseline
Human accessIAM Identity Center only; an SCP denies IAM users and access keysEntra groups only; a policy at mg-acme denies role assignments to individual usersGoogle groups only; IAM members only from your Cloud Identity customer
AuditorsSecurityAuditorAccess and LeadSecurityAuditorAccess with an explicit deny on object, parameter, secret, database and log readsSecurity Reader and acme-security-auditor, which has no data actions; lead auditors add Reader and Log Analytics ReaderSecurity reviewer, Security Command Center and policy viewer roles at the organization; IAM deny policies block object, secret, table and document reads and workload logs
Privileged accessAdministrative permission sets for the two lead groupsWith enable_pim, Owner through PIM elevation groups: at most 12 hours, MFA, justification, approval for Owner on prodOwner through Privileged Access Manager: at most 12 hours, justification, approval for the wider grants

For the AWS edition, the unused-access analyzer reports roles and permissions unused for 90 days, and the Team and Access Management workbook defines a quarterly access audit. The least privilege page shows the full matrix per edition.

How you verify it​

AWS Enterprise Baseline​

  • Config console in the management account, Conformance packs: BuiltForProd-soc2-baseline deployed to every account except sandbox.
  • IAM Identity Center in the identity account: the two auditor permission sets with their inline deny.

Azure Enterprise Baseline​

  • Policy, Compliance, scope mg-acme: acme-soc2-base and acme-mcsb, with every other acme-* framework assignment you switched on.
  • Policy, Exemptions on acme-plat-sandbox: one Waiver per compliance assignment.

GCP Enterprise Baseline​

  • Logging, Logs-based metrics in acme-core-audit: the 13 detection metrics.
  • IAM, Deny policies: acme-audit-protection, acme-sa-protection and acme-auditor-no-data at the organization, acme-auditor-no-workload-logs on the plat folder.

The security posture summary condenses this page and the Secure page for a reviewer; the policy as code page explains why controls expressed in code produce evidence automatically.