Validation & CIS audit · 05

STEP 5 — Post-Build Validation & CIS Benchmark Audit

Run after every landing-zone build and after every workload onboarding wave; then monthly as Day-2 hygiene. Two passes: functional validation (does it work as designed) and compliance audit (is it configured per the benchmark).

5.1 Functional Validation Pass

  • Identity: federated login per persona group; access certification list matches the group registry; break-glass sealed and untested-in-anger.
  • Network: on-prem ↔ hub ↔ spoke traces; DNS resolution both directions; FastConnect/VPN failover test if dual paths exist; NSG default-deny spot checks.
  • Integrations: OIC ↔ Fusion pod round-trip on a test integration per environment; agent heartbeat; SFTP/File Server transfer test; ATP private connectivity.
  • Operations: alarms fire on synthetic breach; events → notifications delivery; flow logs and audit logs queryable; budget alert test at a temporary low threshold.
  • DR posture (if secondary region enabled): replication targets exist (Object Storage replication policies, ATP cross-region backup/ADG per design); backup restore test for one representative resource; region-pair documented RTO/RPO vs questionnaire targets.
  • 5.2 CIS OCI Foundations Benchmark Audit (manual)

    Audit against the CIS Oracle Cloud Infrastructure Foundations Benchmark (current version, Level 1 minimum; Level 2 where the governance standard demands). Method: for each control family, run the documented read-only CLI/console checks, record evidence, log findings as SEC-XX items with severity and owner:

    FamilyRepresentative manual checks
    IAMNo API keys on tenancy admins; MFA enforced; password/lockout policy; no policies granting manage all-resources in tenancy beyond break-glass; unused credentials aged out
    NetworkingNo security lists/NSGs allowing 0.0.0.0/0 to 22/3389; default security lists locked; no public IPs where the exposure rules forbid them
    Logging & MonitoringAudit retention at maximum; Events rules exist for IAM/network changes; VCN flow logs enabled on every subnet in scope; notifications topics subscribed and confirmed
    Object StorageNo public buckets (or documented exceptions); versioning on state and evidence buckets; lifecycle rules where mandated
    Encryption & KeysCustomer-managed keys for mandated data classes; key rotation policy; secrets in Vault, none in code or tfvars
    Cloud Guard / PostureEnabled at tenancy scope; reporting region as decided; problems triaged; responder posture per environment maturity

    Close the loop: every SEC-XX finding gets a remediation MR or a Governance-Board-approved exception with expiry. The audit is complete when the findings register is empty or fully dispositioned — and the signed audit report is filed with the build evidence.

    STEP 5 · DOCUMENT 5 OF 6

    Post-Build Validation & CIS Benchmark Audit

    Consolidated validation after every landing-zone phase and every workload build · CIS OCI Foundations Benchmark audit procedure · findings register · remediation workflow

    Run (a) at each phase gate of Doc 3, (b) after every Doc 4 workload acceptance, and (c) as a full audit before production go-live and quarterly thereafter.

    Post-Build Validation & CIS Audit Runbook

    FieldValue
    Document ID<CUSTOMER>-RUN-VAL-005
    Versionv1.0 (Generic Template)
    Execution modelOPTION A — FULLY MANUAL (human engineers, OCI Console; no IaC/automation in the build path)
    Customer<CUSTOMER> — replace on engagement start
    Tenancy<TENANCY-NAME> / OCID: <TENANCY-OCID>
    Prepared byCloud Platform Engineering
    Review cyclePer engagement; template annually

    1. Validation Model

    TriggerScopeOutput
    Phase gate (Doc 3 G-A1…G-A5)Gate test set for that phase (§2)Gate record: pass/fail per test, evidence links, sign-off
    Workload acceptance (Doc 4 §10)Workload test set + regression of touched shared servicesAcceptance record
    Pre-go-live auditFull §2 + §3 CIS audit + §4 evidence packAudit report + findings register
    Quarterly operations audit§3 CIS audit + access reviews + tag complianceQuarterly compliance report (Doc 1 §14 metrics)
    MANUAL-MODE RULE:Validation tooling is read-only: console inspection, Cloud Guard, and the CIS compliance checker script (§3.2) which only lists and reads resources. Nothing in this document creates, modifies, or deletes a resource — remediations found here go back through the change process and the relevant runbook section.

    2. Consolidated Validation Test Matrix

    2.1 Identity and access

    IDTestPass criterion
    T-ID-01Federated pilot user Console loginIdP redirect + MFA challenged; group-scoped visibility only
    T-ID-02Break-glass login BG-OCI-ADMIN-01ALM-BREAKGLASS-LOGIN fires ≤ 5 min; login recorded in audit
    T-ID-03OCI-DEVELOPER member attempts read in Applications:PRDNot authorized
    T-ID-04Vendor member lists another vendor’s compartmentNot authorized
    T-ID-05OCI-AUDITOR attempts any createNot authorized
    T-ID-06Local-user sweep in all domainsOnly the two break-glass users exist locally
    T-ID-07Service-account register vs VaultEvery service credential exists as a Vault secret with owner + rotation date

    2.2 Structure and governance

    IDTestPass criterion
    T-GV-01Compartment tree read-only export vs design (oci iam compartment list --compartment-id-in-subtree true)Zero drift
    T-GV-02Create test resource without CostCenter tagCreation fails (tag default enforcement)
    T-GV-03Tag compliance sweep (console/Tenancy Explorer or CD3 extract)≥ 98% resources fully tagged; gaps listed with owners
    T-GV-04Budget alert simulation (temporary low-threshold budget)Alert e-mail received by DL; test budget deleted after, logged
    T-GV-05GPU shape launch in Sandbox/DEVBlocked by quota policy
    T-GV-06Policy inventory reviewOne policy per compartment per purpose; no root grants beyond baseline; every statement traceable to Doc 1 §7

    2.3 Network

    IDTestPass criterion
    T-NW-01ATP private endpoint reachability1522 connects from App-Processing subnet with wallet; blocked from Integration-Shared subnet
    T-NW-02OSN reachability via Service Gateway from each subnetObject Storage/OSN endpoints resolve and connect without NAT/IGW
    T-NW-03Ingress sweepNo 0.0.0.0/0 ingress rule anywhere (NSGs + security lists); Cloud Guard clean on the detector
    T-NW-04IGW presence check TST/PRDNo Internet Gateway exists; creation attempt in PRD blocked by Security Zone
    T-NW-05Hybrid (if built): tunnel/BGP state + failoverBoth IPSec tunnels UP; FastConnect BGP Established; failover drill passes in window
    T-NW-06DNS resolution matrixPrivate zones resolve per design; no unapproved forwarders

    2.4 Platform workloads (run per workload, from Doc 4 §10)

    IDTestPass criterion
    T-WL-01End-to-end integration with Vault-sourced credentialsSucceeds; zero inline credentials
    T-WL-02Forced fault → error queue → alert routingAlert to correct SupportGroup destination
    T-WL-03Manual promotion with config planNo hand-edits; four-eyes recorded; checksums match
    T-WL-04SFTP/File Server counterparty exchangeKey-only, chrooted, logged
    T-WL-05Backup evidenceATP backups listed per policy; bucket versioning/lifecycle active; PRD long-term retention shown
    T-WL-06Restore drill (pre-go-live at minimum)ATP point-in-time restore to a clone succeeds; documented duration vs RTO

    3. CIS OCI Foundations Benchmark Audit

    3.1 Scope and method

    The audit assesses the tenancy against the current CIS Oracle Cloud Infrastructure Foundations Benchmark (use the latest published version; record which version and level — Level 1 minimum, Level 2 where the compliance frame demands). Three evidence sources are combined:

  • Console inspection — the recommendation-by-recommendation walkthrough in §3.3, for controls that need human judgment (policy wording, key custody, review cadences).
  • Cloud Guard — many CIS checks map to Cloud Guard detectors; export the problems list filtered to the audit window as machine evidence.
  • CIS compliance checker script — Oracle’s published cis_reports.py (read-only; lists/gets only) run against the tenancy for a complete per-recommendation CSV. Permitted under the manual-only rule as evidence tooling.
  • 3.2 Running the compliance checker (read-only)

  • Run under an OCI-AUDITOR federated identity (read-only by policy — the script cannot mutate anything even if it tried). Use CloudShell (Menu > Developer tools > Cloud Shell) so no laptop credentials are created.
  • Download the current script from Oracle’s official compliance-checker repository (record the commit/version in the audit report).
  • Execute: python3 cis_reports.py -t <profile-or-instance-auth> --report-directory cis_audit_<DATE> (add --all-resources for the full inventory annex).
  • Archive the output CSVs/HTML into the audit evidence bucket or the engagement documentation store; never edit the raw outputs.
  • 3.3 Recommendation walkthrough (summary map)

    The full recommendation list lives in the benchmark itself; the map below shows where each CIS section is evidenced in this engagement and the design intent it should find. Audit = confirm reality matches; deviations go to the findings register (§4).

    CIS sectionWhat it checksWhere this series set itTypical manual-build risks to look for
    1. IAMMFA, no tenancy-admin sprawl, policy hygiene, credential/key rotation, no local users, API key ageDoc 3 A1/A3; Doc 1 §5–7Leftover temporary admin; engineer API keys created during build and never deleted; missing rotation dates on service accounts
    2. NetworkingNo unrestricted ingress (22/3389/0.0.0.0/0), security list vs NSG hygiene, no public IPs where private sufficesDoc 3 A4Temporary test rules left open after connectivity debugging; test VM from VPN validation not deleted
    3. ComputeInstance metadata service v2, no public IPs, patching postureDoc 4 §5.4/§6.2Agent/SFTP VMs built quickly without IMDSv2 enforced
    4. Logging & MonitoringAudit retention 365d, VCN flow logs, object storage write logs, alarms/events for IAM changes, Cloud Guard enabledDoc 2 §6; Doc 3 A4/A5Flow logs enabled on some subnets only; missing event rules for IAM changes (add them if the benchmark version requires — via change process)
    5. Object StorageNo public buckets, versioning, CMK encryption where requiredDoc 4 §4.2Migration-era buckets made public "temporarily"; versioning skipped on ad hoc buckets
    6. Asset ManagementResources tagged and inventoriedDoc 1 §9; Doc 6 registerTag gaps on manually created resources (T-GV-03)
    WARNING:Manual builds fail CIS audits in predictable places: forgotten temporary access, debug-era network rules, and untracked test resources. Sweep the build log for every "temporary" entry and verify each was reverted BEFORE running the audit — finding them yourself is a build-log line; the auditor finding them is a finding.

    4. Findings Register and Remediation

    4.1 Findings register (one row per deviation)

    FieldContent
    IDFND-<YYYY>-<seq>
    SourceCIS <rec #> / Cloud Guard problem ID / test ID (§2)
    SeverityCritical / High / Medium / Low (CIS profile level + data classification of affected resource)
    Resource(s)OCID(s) + compartment
    Description & evidenceWhat deviates from which control; screenshot/CSV reference
    OwnerPer control-family ownership (Doc 1 §13)
    Remediation pathRunbook section that rebuilds it correctly (e.g., Doc 3 §5.3 NSG rules) + change record ID
    Due / closedSLA per severity: Critical same-day, High 7d, Medium 30d, Low next quarter

    4.2 Remediation workflow

  • Triage findings with Security & Compliance within 24h of the audit; agree severity and owner per row.
  • Raise one change record per remediation (or one batched record per control family for low-risk items).
  • Execute remediations via the original runbook sections — never ad hoc console fixes outside the documented procedures (that recreates the drift the audit exists to catch).
  • Re-test the specific control after each remediation; attach re-test evidence to the finding; close.
  • Exceptions the customer accepts instead of remediating go to the exception register with an approver, an expiry date, and a compensating control — never an open-ended acceptance.
  • Report the CIS score trend (findings by severity, open vs closed) in the quarterly governance report (Doc 1 §14).
  • 5. Go-Live Readiness Gate

    #Criterion
    1All §2 validation tests green, evidence archived.
    2CIS audit complete; zero open Critical findings; High findings remediated or formally excepted with compensating controls.
    3Cloud Guard: no open Critical problems in PRD/Analytics/Regulated; detector recipes at agreed tuning.
    4Restore drill (T-WL-06) completed within RTO; backup evidence archived.
    5Break-glass procedure tested this quarter; credentials custody confirmed.
    6Access review completed: group memberships, vendor time-boxes, service-account register.
    7As-built documentation current: CD3 extract reconciled, diagrams updated (Doc 6), build log complete.
    8O&M handover accepted: alarm destinations verified, on-call rota published, runbook set delivered.

    Gate approved by the Cloud Governance Board → production cutover may proceed under the customer cutover plan.

    Revision History

    VersionDateDescription
    1.0<DATE>Generic template — validation + CIS audit edition.