Join our Newsletter — 33% off our NHI Course
Home› Guides› How to Audit Your Cloud Identity: 11 Health…
Template Cloud & Workload Identity

How to Audit Your Cloud Identity: 11 Health Check Metrics for IAM Drift

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 13 November 2026 11 min read
On this page

Cloud permissions drift between reviews: a clean audit in January says little about the keys, roles and trusts that exist in June. This health check turns cloud identity hygiene into 11 measured indicators that you track every month and sign off every quarter, so drift shows up as a trend line rather than as a surprise in an incident. It complements the one-off Cloud IAM Review Checklist, which tells you how to run a full review and what each control should look like; this page assumes those controls exist and asks whether they are holding.

It is written for cloud security teams, identity architects and internal audit across any public cloud platform or combination of platforms. The metrics cover human and non-human identities alike, because in most cloud estates roles, service accounts and access keys outnumber people many times over. For the design of the controls themselves, see the Cloud PAM and CIEM Guide.

What this health check covers

Scope is every cloud account, subscription and project in the organisation, plus the identity provider that federates people into them. It measures the identity control plane: root and emergency accounts, people, workload identities, long-lived credentials, policies, cross-account and external trust, and the gap between granted and used permissions. It does not measure workload configuration, network exposure or data stores; those belong to posture management and the Cloud Security Posture Management (CSPM) Guide.

  • Who runs it: the cloud security or cloud platform team collects the data; the identity team owns federation metrics; internal audit samples the evidence.
  • How it differs from the review checklist: the checklist is a point-in-time pass or fail against each control. The health check is a repeatable set of ratios with thresholds, read as a trend over at least three periods.
  • What it catches: keys that quietly age past policy, wildcard policies that creep back in after a clean-up, new third-party trusts nobody reviewed, people who bypass single sign-on with local console users, and roles that keep permissions they never use.

Key takeaways

  • Measure ratios, not counts. "37 old keys" means nothing without the denominator; a share of all keys shows whether the estate is getting better as it grows.
  • Non-human identities drive most metrics. Access keys, workload roles and cross-account trusts change far more often than human admin rights, so give them weekly or monthly collection.
  • The permission utilisation gap is the drift indicator. It shows how much granted access goes unused across the usage window, and it rises long before an incident makes it visible.
  • Any root or emergency account use without a ticket is red. No trend analysis is needed for that metric; investigate each event.
  • Thresholds are starting points. Tighten them each year as automation, federation and just-in-time access mature.

How to use this template

Copy the metrics table and the findings log into your own tracker, then run this Cloud Identity Health Check at the cadence set out below. Collect each figure from the data sources described, score it against the green, amber and red thresholds, and record every amber or red result as a finding with an owner and a due date. Keep each completed review so you can compare results quarter on quarter: the trend matters as much as any single score.

Cadence and sign-off

ReviewFrequencyOwnerApproverEvidence kept
Root and emergency account use (CI-01)Continuous alert, weekly checkCloud security operationsHead of cloud securityAlert record, matching change or incident ticket
Credential metrics (CI-02, CI-03, CI-07)MonthlyCloud platform teamCloud security leadDated credential and key inventory export
Policy and permission metrics (CI-04, CI-05, CI-10)MonthlyCloud security teamCloud security leadPolicy export, unused access findings with run date
Trust and privilege metrics (CI-06, CI-08, CI-09, CI-11)QuarterlyCloud security team with identity teamCISO or delegateTrust relationship list, admin role holder list, sponsor confirmations
Full health check report and trendQuarterlyCloud security leadCISO; risk committee for red metricsMetric table for three periods, findings log, signed approval
Independent sampleAnnuallyInternal auditAudit committeeRe-performed calculation for at least three metrics

Metrics

All metrics are calculated per platform and for the estate as a whole. Use a 90-day usage window unless stated. NHI Mgmt Group recommended thresholds; adjust to your risk appetite.

IDMetricFormula (numerator / denominator)Data source typeGreenAmberRed
CI-01Root account useRoot or emergency account sign-ins and API calls without a linked ticket in the period (count, no denominator)Cloud control-plane audit logs, alert history0Not used (any unticketed use is red)1 or more
CI-02Long-lived access key shareActive workloads authenticating with a static access key or service account key / all active workload identitiesCredential report, key inventory, workload identity listUnder 10%10 to 25%Over 25%
CI-03Key age over 90 daysActive static keys older than 90 days / all active static keysCredential report with key creation datesUnder 5%5 to 15%Over 15%
CI-04Unused role rateRoles, service accounts and users with no authentication in 90 days / all roles, service accounts and usersLast-used data, unused access findingsUnder 5%5 to 15%Over 15%
CI-05Wildcard policy shareCustom policies or role definitions granting all actions, all resources, or a whole service with write rights, outside approved admin roles / all custom policiesPolicy and role definition exportUnder 2%2 to 5%Over 5%
CI-06Cross-account trusts unreviewedCross-account, cross-tenant and workload federation trusts without a sponsor confirmation in the last 12 months / all such trustsTrust policy export, external access findings, sponsor register0%Above 0% to 10%Over 10%
CI-07Workload federation shareWorkloads outside the cloud (CI pipelines, other clouds, on-premises) using federated short-lived tokens / all workloads outside the cloud that call cloud APIsFederation provider config, CI job identity settings, key inventoryOver 80%50 to 80%Under 50%
CI-08Human console users without SSOPeople with local cloud console passwords or local users / all people with console accessCredential report, identity provider assignment listOnly documented emergency accountsAbove 0% to 2% beyond themOver 2% beyond them
CI-09Admin role holdersIdentities with standing administrator or access-granting roles / all identities with any role; track human and non-human separatelyRole assignment export including inherited assignments, just-in-time eligibility listUnder 2% and all humans just-in-time2 to 5% or some standing human adminsOver 5% or growing three periods running
CI-10Permission utilisation gapGranted permissions (actions or services) not used in 90 days / all granted permissions, averaged across rolesUnused permission findings, last-accessed data, role recommendation outputUnder 30%30 to 60%Over 60%
CI-11External principal grantsGrants to principals outside the organisation (guests, foreign accounts, public or anonymous access) without a sponsor and end date / all external principal grantsExternal access findings, guest user list, resource policy export0%Above 0% to 5%Over 5%

Reading the metrics together

  • CI-02 and CI-07 move together: as workload federation rises, the long-lived key share should fall. If federation rises and key share does not, old keys were left active after migration. See the Workload Identity Federation Guide.
  • CI-04 and CI-10 are different signals: unused roles are identities to delete; the utilisation gap is excess permission on identities that are in use and need right-sizing.
  • CI-09 needs a growth check: a stable share hides absolute growth in a fast-growing estate, so report the count alongside the ratio.
  • CI-06 and CI-11 cover different doors: trusts let another account assume a role; grants put an outside principal straight onto a resource or role.

How to collect the data

Pull every figure from the platform itself, not from documentation or a configuration management record, and store each export with its collection time. The same export sources serve every period, which is what makes the trend credible.

  1. Credential inventory: the cloud platform's credential report or key listing per account, with key creation date, last used date and owning identity. This feeds CI-02, CI-03 and CI-08.
  2. Identity and role inventory: a full export of users, roles, service accounts and managed identities with last authentication time, from the cloud identity service across all accounts in the organisation. This feeds CI-04 and the denominators of CI-02 and CI-09.
  3. Policy and role definitions: every custom policy and role definition at organisation, account and resource scope. Parse them for wildcard actions and resources rather than relying on names; this feeds CI-05.
  4. Unused access and permission analysis: the platform's native unused access analyser or role recommender output, or your own analysis of control-plane audit logs over the 90-day window. Record whether the analyser covers the full window, because some native tools look back less far than others. This feeds CI-04 and CI-10.
  5. Trust and external access: role trust policies, cross-tenant settings, workload federation configuration and external access findings, matched against a sponsor register held by the cloud security team. This feeds CI-06 and CI-11.
  6. Identity provider records: application assignments and group membership for the cloud federation connection, and sign-in logs, to show who reaches the console through single sign-on. This feeds CI-08 and the human part of CI-09.
  7. Control-plane audit logs: organisation-wide, retained centrally, filtered for root and emergency account events. Match each event to a change or incident ticket for CI-01.

Automate collection on a schedule using a dedicated read-only audit identity per platform. That identity is itself privileged reach into every account: give it read-only rights, federated credentials and its own entry in the trust register.

Red flags that need action now

  • Any root or emergency account sign-in or API call with no matching ticket.
  • A static key over 365 days old on an identity with write or administrator rights.
  • A new policy granting all actions on all resources outside the approved admin role set since the last period.
  • A cross-account trust to an account that no current supplier or business unit owns.
  • Public or anonymous access granted through a resource policy on a store or key that holds sensitive data.
  • A people account with console access and no single sign-on, created after the last review.
  • A workload federation trust whose subject condition accepts any repository, branch or audience. See the Workload Identity Federation Trust Review Checklist.
  • The admin role holder count up by more than 20% in one quarter without a matching change programme.
  • Collection failed for any account, so the metric covers less than the full estate.

Findings log

Record one row per finding. Use the metric ID so the finding can be traced back to the trend; close a finding only when the next collection shows the metric has moved.

Finding IDFindingMetricSeverityOwnerDue dateStatus
HC-CLOUDID-01[Example: 14 static keys over 90 days on CI pipeline identities in production accounts]CI-03[High][Platform engineering lead][DD/MM/YYYY][Open]
HC-CLOUDID-02[Example: 3 supplier trusts with no sponsor confirmation]CI-06[Medium][Vendor management][DD/MM/YYYY][Open]
HC-CLOUDID-03[Finding][CI-xx][Critical, High, Medium, Low][Named owner][DD/MM/YYYY][Open, In progress, Closed, Risk accepted]

Suggested service levels: critical within 7 days, high within 30 days, medium within 90 days. A risk acceptance needs an expiry date and the approver from the cadence table.

Remediation map

Finding typeMetricsWhere the fix is described
Root or emergency account misuse or weak controlsCI-01Break-Glass Account Guide; Protecting Cloud Control-Plane Recovery Accounts and Credentials
Old or unnecessary static keysCI-02, CI-03Secrets Rotation Runbook Template; Cloud Access Key Leak Playbook if exposure is suspected
Workloads outside the cloud still on keysCI-07Workload Identity Federation Guide; Cloud Workload Identity Guide
Unused identities and excess permissionsCI-04, CI-10Least Privilege Techniques Guide; Cloud PAM and CIEM Guide
Wildcard and admin-equivalent policies, standing adminsCI-05, CI-09Cloud IAM Review Checklist, sections 6 to 9; Privileged Access Management Guide
Unreviewed trusts and external grantsCI-06, CI-11Workload Identity Federation Trust Review Checklist; Access Reviews Guide
People bypassing single sign-onCI-08Cloud IAM Review Checklist, human access checks

References

Related NHI Mgmt Group resources: Cloud IAM Review Checklist · Cloud PAM and CIEM Guide · Cloud Identity and CIEM Maturity Model · Workload Identity Federation Guide

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 13 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org