Most AWS estates cannot answer two basic questions: which resources can be reached from outside the organisation, and which roles and keys hold permissions they never use. AWS IAM Access Analyzer is the native service built to answer both, and it also validates policies before they are deployed. This guide explains its three analyzer types (external access, internal access and unused access), zones of trust, regional behaviour, archive rules, EventBridge events, policy validation and custom policy checks, and how to run it across an AWS Organization from a delegated administrator account. It goes deeper than our Cloud IAM Review Checklist and is written for AWS security engineers, IAM owners and platform teams.
This is an independent guide from NHI Mgmt Group, based on AWS's public documentation as of November 2026. It is not sponsored or reviewed by AWS. Product features change, so confirm settings against AWS's current documentation.
Key takeaways
- Three analyzers, three questions. External access finds resources shared outside your zone of trust; internal access shows which principals inside the organisation can reach selected critical resources; unused access finds roles, access keys, passwords and permissions with no recent use.
- External access is regional. AWS states that external access analyzers only analyse policies in their own region, so you need one in every region where you have supported resources. Unused access is not region-dependent.
- External findings are free; the others are not. Unused access is charged per IAM role and user analysed per analyzer per month, internal access per resource monitored, and custom policy checks per API request.
- Shift left with custom policy checks. CheckNoNewAccess, CheckAccessNotGranted and CheckNoPublicAccess can fail a pipeline before an over-permissive policy reaches production.
- Archive rules are a security control. A careless rule hides real exposure. Keep the public access filter set to false so archive rules never suppress public access findings.
How identity and access analysis works in IAM Access Analyzer
When you create an analyzer you choose a zone of trust: a single account or your entire organisation. Principals inside the zone are trusted; access granted to anything outside it produces an external access finding. AWS states that new or changed policies are typically analysed within about 30 minutes, with delays of up to 6 hours for S3 multi-Region access points and up to 24 hours where CloudTrail delivery or resource control policy changes do not trigger a rescan. After a fix you can confirm with Rescan in the console or the StartResourceScan API.
| Analyzer | Question it answers | Scope | Regional? | Cost |
|---|---|---|---|---|
| External access | Which resources are shared with another account, root user, IAM user or role, federated user or anonymous user outside the zone of trust? | Resource-based policies on 15 resource types | Yes: one analyzer per region | Free |
| Internal access | Which principals inside my organisation or account can reach this business-critical resource? | Selected resources on 6 resource types | Not stated by AWS on the overview page; check per region | Per resource monitored per analyzer per month |
| Unused access | Which roles, access keys, passwords, services and actions have not been used in the tracking period? | All IAM roles and users | No | Per IAM role and user analysed per analyzer per month (service-linked roles excluded) |
Supported resource types. External access covers S3 buckets and directory buckets, IAM roles (trust policies), KMS keys, Lambda functions and layers, SQS queues, Secrets Manager secrets, SNS topics, EBS volume snapshots, RDS DB snapshots and DB cluster snapshots, ECR repositories, EFS file systems, and DynamoDB streams and tables. Internal access covers the subset of S3 buckets and directory buckets, RDS DB and cluster snapshots, and DynamoDB streams and tables. Resources not in these lists, and access granted through mechanisms other than resource-based policies, are outside what the external analyzer sees.
For the NHI view, the IAM role entry matters most: an external access finding on a role means its trust policy lets a principal outside your zone assume it. This is the cross-account trust that vendor integrations, CI systems and SaaS tools rely on. See the AWS STS and AssumeRole Guide and the Cross-Account Access Guide.
Admin roles and privileged access
Access Analyzer itself needs governance: whoever controls analyzers and archive rules decides what the security team sees.
- Delegated administrator: run organisation analyzers from a delegated administrator member account (typically the security tooling account), not the management account. Only the management account can add, remove or change the delegated administrator. AWS warns that changing it disables the organisation analyzers the former administrator created; their findings become inaccessible until the account is re-registered, and the new administrator must create new analyzers.
- Member accounts: an account that is neither the management account nor a delegated administrator can create only an analyzer scoped to itself. That is useful for teams, but the organisation view must live in the security account.
- Service-linked role: creating an analyzer creates AWSServiceRoleForAccessAnalyzer in the account. Do not block it with SCPs.
- Protect the analyzers: use an SCP to deny access-analyzer:DeleteAnalyzer, access-analyzer:CreateArchiveRule, access-analyzer:UpdateArchiveRule and access-analyzer:UpdateFindings to everyone except the security team's role. An attacker who can archive findings can hide a public bucket they just created.
- Cost of duplicates: an unused access analyzer in both the management account and the delegated administrator account is charged twice. Keep one organisation-level unused analyzer.
See the AWS Organizations and SCPs Guide and the AWS Multi-Account Security Reference Architecture Guide.
Integrations, API tokens and service accounts
Unused access for non-human identities
Unused access is the most useful Access Analyzer feature for NHI hygiene. It reports four kinds of finding: unused roles (no access activity in the window), unused IAM user access keys, unused IAM user passwords, and unused permissions (service-level and, for a listed set of services, action-level permissions an active role or user did not use). AWS states that analyzers evaluate identity-based policies attached to roles for unused permissions.
- Tracking period: configurable from 1 to 365 days. Only entities that existed for the whole period are evaluated, so a new role will not be flagged until it is older than the window. A 90-day window is a common starting point; quarterly or annual jobs (disaster recovery, year-end processing) need longer, or a tag exclusion.
- Exclusions: organisation analyzers can exclude account IDs; both account and organisation analyzers can exclude IAM users and roles by tag (key 1 to 128 characters, not starting with aws:). Accounts added to an OU later are not excluded automatically. Use exclusions sparingly and review them, because an excluded break-glass role is also an unmonitored one.
- Removed accounts: if a member account leaves the organisation, findings stop updating after 24 hours and are removed after 90 days.
| Finding | Typical NHI cause | Remediation |
|---|---|---|
| Unused role | Retired integration, abandoned proof of concept, vendor role left after offboarding | Confirm owner, remove trust policy first (reversible), then delete |
| Unused access key | Key issued to a script or tool that moved to role-based access | Deactivate, wait for errors, then delete; replace long-lived keys with roles |
| Unused password | IAM user created for a service but given console access | Remove the login profile; service users need no console password |
| Unused permissions | Role granted a managed policy far broader than its job | Generate a policy from CloudTrail activity and replace the broad policy |
See the AWS Access Keys Security Guide.
Policy generation
Policy generation analyses CloudTrail logs for the actions and services an IAM user or role used in a date range you choose and produces a policy you can attach. It is the practical way to right-size service roles. Choose a range that covers the role's full business cycle, review the generated policy (it reflects what happened, which may include mistakes), and test in a non-production environment before replacing the live policy.
Policy validation and custom policy checks
Policy validation checks policies against IAM policy grammar and AWS best practices, returning security warnings, errors, general warnings and suggestions. It runs in the IAM console JSON editor, the CLI and the API. Custom policy checks compare policies with your own standards:
| Check | CLI command | Use in a pipeline |
|---|---|---|
| CheckNoNewAccess | check-no-new-access | Fail when an updated policy grants access the existing or reference policy did not |
| CheckAccessNotGranted | check-access-not-granted | Fail when a policy allows actions you list as critical, for example iam:PassRole or kms:Decrypt on production keys |
| CheckNoPublicAccess | check-no-public-access | Fail when a resource policy could grant public access for a given resource type |
The open source IAM Policy Validator for AWS CloudFormation (cfn-policy-validator, from AWS Labs) parses policies from CloudFormation templates and sends them to Access Analyzer. Its validate command fails the build on ERROR and SECURITY_WARNING findings by default, and the custom checks fail on any finding unless --treat-findings-as-non-blocking is set. Each custom policy check call is charged. See the AWS CloudFormation and CDK Security Guide.
Logging and detection
Access Analyzer publishes events to Amazon EventBridge with source aws.access-analyzer. AWS states events arrive typically within about an hour of the occurrence.
| detail-type | Emitted for | Suggested routing |
|---|---|---|
| Access Analyzer Finding | External access findings: new, status change, deletion | Status ACTIVE and public access true: page the on-call team; others: ticket to resource owner |
| Internal Access Finding | Internal access findings | Data owner review queue |
| Unused Access Finding for IAM entities | Unused roles, keys, passwords and permissions | IAM or NHI owner backlog with due dates |
| Access Preview State Change | Access preview CREATING, COMPLETED, FAILED | Pipeline feedback |
Deleted findings keep their status and set isDeleted to true, so detection logic must check that field. Findings also flow to AWS Security Hub CSPM, which is the simplest way to aggregate across regions. Alongside findings, watch CloudTrail for changes to the analyzers themselves: DeleteAnalyzer, CreateArchiveRule, UpdateArchiveRule and UpdateFindings calls by anyone other than the security team. See the AWS Security Hub Guide.
Archive rules. Archive rules automatically archive new findings that match criteria such as resource type, resource owner account or principal, with up to 20 values per criterion. By default they apply only to new findings; "Create and archive active findings" also archives matching existing findings. Rules are regional, so changes must be repeated in each region. AWS does not validate filter values, so a mistyped account ID silently matches nothing or the wrong thing. Use archive rules for intended, documented sharing (a central logging bucket, a known partner account) and always add Public access equals false.
Metrics:
- Regions with an active external access analyzer as a share of regions in use, target 100%.
- Active external access findings older than 14 days, and public access findings older than 24 hours.
- Unused roles and unused access keys per account, trending down quarter on quarter.
- Percentage of NHI roles whose policy was generated or reviewed from activity in the last 12 months.
- Pipeline policy changes blocked by custom policy checks, and the share later approved as exceptions.
- Number of archive rules, each with an owner and a documented reason.
Hardening baseline
- Delegate: register the security tooling account as delegated administrator from the management account.
- External access everywhere: create an organisation-scoped external access analyzer in every enabled region, ideally through CloudFormation StackSets or your landing zone tooling so new regions are covered.
- Triage the backlog: sort findings by public access first, then cross-organisation role trusts, then KMS keys and secrets. Fix or document each. Create archive rules only for documented intended access.
- Unused access: create one organisation unused access analyzer with a 90-day window. Pilot remediation in one account: deactivate keys and remove role trust before deleting, so rollback is a re-activation.
- Internal access: enable internal access only for named crown-jewel resources (the S3 buckets, DynamoDB tables and RDS snapshots holding regulated data), because it is charged per resource.
- Shift left: add validate and CheckNoNewAccess or CheckAccessNotGranted to infrastructure pipelines in non-blocking mode for two weeks, review the noise, then switch to blocking.
- Route events: EventBridge rules for each detail-type, plus CloudTrail alerts on analyzer and archive rule changes.
Small versus large estates: a single-account team can run account-scoped analyzers, review findings weekly in the console and still benefit from free external findings. A large organisation needs the delegated model, regional coverage through automation, ticket routing by account owner tag, and a policy for exclusions and archive rules approved by the security team.
Framework mapping
| Capability | NIST SP 800-53 Rev 5 | NIST CSF 2.0 | CIS Controls v8.1 |
|---|---|---|---|
| Unused access findings and removal | AC-2(3) Disable Accounts; AC-6(7) Review of User Privileges | PR.AA-05 | 5.3 Disable Dormant Accounts |
| Policy generation and right-sizing | AC-6 Least Privilege | PR.AA-05 | 6.8 Define and Maintain Role-Based Access Control |
| External and public access findings | AC-3 Access Enforcement; AC-21 Information Sharing | ID.RA-01; DE.CM-09 | 3.3 Configure Data Access Control Lists |
| Policy validation and custom checks in CI | CM-3 Configuration Change Control; SA-11 Developer Testing and Evaluation | PR.PS-01 | 16 Application Software Security |
The threat these address is abuse of valid cloud credentials and trust: MITRE ATT&CK T1078.004 Cloud Accounts and T1199 Trusted Relationship. In ISO/IEC 27001:2022, the closest controls are A.5.15 Access control, A.5.18 Access rights and A.8.2 Privileged access rights.
Common misconfigurations
- One region only: an external access analyzer in the home region while resources sit in others, so a public snapshot elsewhere is never reported.
- Over-broad archive rules: a rule archiving all findings for a resource type or a whole account, applied with "archive active findings", which silently clears real exposure.
- Treating findings as noise: large external access backlogs nobody owns, often vendor role trusts that outlived the contract.
- Deleting before disabling: removing an "unused" role or key outright, breaking an annual job; deactivate and remove trust first, delete later.
- Blind trust in generated policies: attaching a generated policy without review, which also codifies whatever the role did wrong during the window.
- Exclusions without review: tag exclusions that grow until the most privileged roles are no longer analysed.
Practitioner checklist
- Register a delegated administrator for IAM Access Analyzer in the security tooling account.
- Deploy an organisation external access analyzer in every region in use, automatically for new regions.
- Create one organisation unused access analyzer and set a tracking period that fits your business cycles.
- Route EventBridge findings by detail-type to owners, and page on active public access findings.
- Restrict who can delete analyzers, edit archive rules or update findings, and alert on those calls.
- Keep Public access equals false in every archive rule and document each rule's reason and owner.
- Right-size NHI roles using policy generation, tested before production.
- Add policy validation and custom policy checks to infrastructure pipelines, moving from non-blocking to blocking.
- Enable internal access analysis for named crown-jewel data stores.
Standards and references
- AWS: Using AWS Identity and Access Management Access Analyzer
- AWS: IAM Access Analyzer supported resource types
- AWS: Creating an unused access analyzer
- AWS: IAM Access Analyzer custom policy checks
- AWS: Archive rules
- AWS: Monitoring IAM Access Analyzer with Amazon EventBridge
- AWS: Delegated administrator for IAM Access Analyzer
- AWS Labs: IAM Policy Validator for AWS CloudFormation
Related NHI Mgmt Group resources: Cloud IAM Review Checklist · AWS IAM Policies and Roles Guide · AWS Resource Control Policies and Data Perimeter Guide · Top 10 AWS IAM Mistakes · AWS Lambda Security Guide