AWS Security Hub only gives an organisation-wide view if it is enabled everywhere, aggregated into one Region, owned by the right account and tuned so that the findings people see are the ones that matter. AWS now offers two related services: AWS Security Hub, the unified service that correlates signals into exposures using the Open Cybersecurity Schema Framework (OCSF), and AWS Security Hub CSPM, the posture service (formerly called Security Hub) that runs standards and controls and uses the AWS Security Finding Format (ASFF). This guide covers both as they stand in October 2026: delegated administration in the Security Tooling account, central configuration, cross-Region aggregation, standards and controls, automation rules, EventBridge response and the IAM permissions that protect the service itself. It is for cloud security engineers and architects running multi-account AWS estates. For the wider AWS picture, see the AWS Security Guide.
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
- Two services, one delegated administrator. Security Hub CSPM produces control findings; Security Hub correlates them with GuardDuty, Inspector, Macie and IAM Access Analyzer into exposures. Designating the CSPM delegated administrator also makes that account the Security Hub delegated administrator.
- Use the Security Tooling account, never the management account. The management account cannot designate itself as delegated administrator, and the AWS Security Reference Architecture puts Security Hub in the Security Tooling (Control Tower Audit) account.
- Central configuration prevents drift. Configuration policies set service state, standards and controls for accounts and OUs across a home Region and linked Regions, and block members from changing them.
- AWS Config is a hard dependency. Most CSPM controls run as service-linked AWS Config rules, so Config must record resources in every account and Region, including the delegated administrator account.
- Tune with controls first, rules second. Disable controls you do not need (this stops checks and charges); use automation rules to raise severity for production or suppress known noise, and route the rest through EventBridge or ticketing.
How identity and access work in AWS Security Hub
Security Hub has no users of its own. Access is IAM permissions in each account, plus an organisation relationship in which one delegated administrator account sees and manages findings for member accounts. That relationship is the core design decision.
The two services
| Aspect | AWS Security Hub | AWS Security Hub CSPM |
|---|---|---|
| Role | Correlates and prioritises signals; exposures and attack paths; response workflows | Posture checks against standards; collects findings from AWS and partner products |
| Finding format | OCSF | ASFF |
| Inputs | Security Hub CSPM, GuardDuty, Inspector, Macie, IAM Access Analyzer | Its own controls (via AWS Config) and integrated products |
| Distinctive features | Exposure findings, attack path graph, unused access analysis (90-day lookback) | Standards, controls, control parameters, insights, central configuration |
| Automation rules | Finding rules (Comment, SeverityId, StatusId, Jira Cloud and ServiceNow ITSM tickets) and resource criticality rules | ASFF rules updating Severity, Workflow, Note, Confidence, Criticality and more |
| Rule limits | 100 per account per Region | 100 per administrator account, per Region |
Account roles in an organisation
- Management account: designates the delegated administrator. It cannot designate itself. Keep its use to that single action.
- Delegated administrator (Security Tooling account): enables and configures members, creates configuration policies and automation rules, and sees all member findings. Align it with the delegated administrators for GuardDuty, Inspector, Macie and Detective; if the Detective administrator differs, pivots from findings into Detective do not work.
- Member accounts: generate findings. Centrally managed members cannot call account-level operations such as
EnableSecurityHub,BatchEnableStandardsorUpdateStandardsControlin governed Regions, but can still read status.
See the AWS Multi-Account Security Reference Architecture Guide for where the Security Tooling and Log Archive accounts sit.
Admin roles and privileged access
Whoever administers Security Hub can make problems disappear: disabling a control, suppressing findings with a rule or removing a member all reduce what defenders see. Treat those permissions as security-critical, in the same tier as the logging pipeline.
| AWS managed policy | What it grants | Give it to |
|---|---|---|
| AWSSecurityHubFullAccess | All Security Hub CSPM actions, plus GuardDuty and Inspector administration actions and creating the service-linked role | A small platform team in the Security Tooling account, through an elevated role |
| AWSSecurityHubReadOnlyAccess | Get, List and Describe; cannot change finding status or configuration | Analysts and auditors by default |
| AWSSecurityHubOrganizationsAccess | Designate the delegated administrator, enable members, related Organizations reads | Management account deployment role only |
| AWSSecurityHubServiceRolePolicy | Service-linked role: reads CloudTrail, CloudWatch, Organizations and tags, manages service-linked Config rules | Attached by AWS; do not modify |
- Split triage from configuration: analysts need to update workflow status on findings (
BatchUpdateFindingsorBatchUpdateFindingsV2) but not to disable controls, delete automation rules or change configuration policies. Build a custom triage role rather than handing out full access. - Guard with SCPs: deny
securityhub:DisableSecurityHub, disassociation from the administrator and AWS Config recorder deletion in member accounts. Changing the delegated administrator should require the management account. See the AWS Organizations and SCPs Guide. - Human access through Identity Center: use permission sets with short sessions for console access to the Security Tooling account. See the AWS IAM Identity Center Guide.
Disabling or tampering with security tooling is a recognised attacker step (MITRE ATT&CK T1685, Disable or Modify Tools), so the guardrails above are detection-preserving controls, not just hygiene.
Integrations, API tokens and service accounts
Security Hub sits in the middle of a set of machine identities: service-linked roles that collect data, automation that updates findings, and connectors that push findings to SIEM, SOAR and ticketing tools.
- Service-linked roles: Security Hub CSPM uses a service-linked role to run checks; enabling Security Hub creates a service-linked IAM Access Analyzer for unused access analysis. Do not try to restrict these with SCPs, or checks fail silently as missing data.
- Ticketing integrations: Security Hub automation rules can open Jira Cloud and ServiceNow ITSM tickets. The integration holds credentials for those platforms; store them with the minimum project or table scope and rotate them. See the API Key Management Guide.
- EventBridge: Security Hub CSPM emits Security Hub Findings - Imported for every new or updated finding, Security Hub Findings - Custom Action when an analyst sends up to 20 findings to a custom action, and Security Hub Insight Results. The administrator account and the aggregation Region receive events for members and linked Regions.
- Remediation functions: Lambda functions or Systems Manager runbooks triggered by findings often hold broad write permissions across accounts. Give each a dedicated role per remediation, restrict which accounts it can assume into, and log every action. See the AWS Systems Manager Security Guide.
- SIEM pull: a SIEM reading findings via API needs only read permissions in the delegated administrator account and aggregation Region, not in every member.
- Identity findings: unused roles, users, access keys and permissions surfaced by the IAM Access Analyzer integration are a direct input to NHI clean-up. See the IAM Access Analyzer Guide and the AWS Access Keys Security Guide.
Logging and detection
Security Hub is a consumer of detection, but its own configuration is also something to monitor. CloudTrail records Security Hub API calls; send the organisation trail to the Log Archive account. See the AWS CloudTrail Guide.
What to alert on
- Coverage loss:
DisableSecurityHub, member disassociation, AWS Config recorder stopped or deleted, standards disabled (BatchDisableStandards). - Silent tuning: new or changed automation rules that set Workflow to SUPPRESSED or lower severity, and controls disabled outside change windows.
- High-value findings: GuardDuty findings for credential misuse, and CSPM failures for controls such as IAM.4 (root user access key should not exist) and EC2.19 (security groups should not allow unrestricted access to ports with high risk).
Aggregation and routing matter because stolen IAM keys with admin-like rights can be put to use within minutes. A GuardDuty finding for credentials used from an unusual network only helps if it reaches an on-call responder in minutes, across every account and Region (T1078.004, T1098.001, T1496.001).
Metrics that show it is working
- Coverage: accounts and Regions with Security Hub CSPM and AWS Config enabled, against the Organizations account list. Target 100% of active Regions.
- Security score trend: FSBP control pass rate per OU, reviewed monthly.
- Mean time to triage: time from CRITICAL or HIGH finding creation to a status change from NEW.
- Exposure closure time: median days to resolve Security Hub exposure findings on production resources.
- Suppression ratio: share of findings suppressed by rules, with each suppression rule owned and reviewed quarterly.
- Drift: number of self-managed accounts outside central configuration. Target zero outside documented exceptions.
Hardening baseline
Standards to enable
| Standard in Security Hub CSPM | Use it when | Notes |
|---|---|---|
| AWS Foundational Security Best Practices (FSBP) v1.0.0 | Always, as the baseline | The recommended configuration policy enables it and all new FSBP controls |
| CIS AWS Foundations Benchmark | You report against CIS | Overlaps FSBP; enabling both raises cost unless consolidated control findings are used |
| NIST SP 800-53 Revision 5 | US federal or FedRAMP-aligned programmes | Maps controls to 800-53 |
| NIST SP 800-171 Revision 2 | Handling Controlled Unclassified Information | Note the standard tracks Revision 2 |
| PCI DSS | Accounts in cardholder data scope only | Apply by OU to in-scope accounts |
| AWS Resource Tagging | Tags drive ownership and criticality | Supports resource criticality rules |
| AI Security Best Practices | Accounts running AI workloads | Network isolation, encryption and KMS checks |
| CRI Profile v2.2 | Financial services | Cyber Risk Institute profile |
| Service-managed standard: AWS Control Tower | You manage detective controls from Control Tower | Configure from Control Tower, not Security Hub |
Rollout
- Prepare: confirm AWS Config is recording in every account and Region, including the Security Tooling account; FSBP Config.1 checks this.
- Delegate: from the management account, designate the Security Tooling account as delegated administrator.
- Choose the home Region: pick a non opt-in Region as home and aggregation Region, and link every Region you operate in. Opt-in Regions must be enabled in accounts before policies apply there.
- Central configuration: start with the recommended policy (Security Hub CSPM on, FSBP with all current and new controls) at the root, then add custom policies per OU (for example, PCI DSS for payment OUs). The limit is 20 custom policies.
- Pilot tuning: run for two weeks on a sandbox OU, disable controls that do not apply (for example, those for services you block by SCP), and set control parameters.
- Automate: add automation rules to raise severity for production accounts and critical resources, and route CRITICAL and HIGH to on-call through EventBridge or ticketing.
- Roll back: configuration policies can be disassociated from an OU, which leaves targets inheriting from the parent; a target can also be switched to self-managed temporarily. Automation rules can be disabled without deletion.
Automation rules that earn their place
Both services evaluate rules only on findings ingested after the rule exists, and neither triggers rules on manual updates through BatchUpdateFindings or BatchUpdateFindingsV2. Rules apply in ascending RuleOrder, so the highest number has the final say when two rules change the same field. With Region aggregation, Security Hub rules are created in the home Region and apply to linked Regions; Security Hub CSPM rules apply only in the Region where they are created, so deploy them to each Region through infrastructure as code.
| Rule | Criteria | Action |
|---|---|---|
| Production uplift | Account ID in the production list, severity HIGH | Raise severity to CRITICAL |
| Crown-jewel resources | Resource tag such as data-classification=restricted | Resource rule: criticality Very High |
| Exploitable vulnerabilities | vulnerabilities.is_exploit_available true | Create Jira Cloud or ServiceNow ticket for the owning team |
| Sandbox noise | Sandbox OU accounts, severity INFORMATIONAL | Set status to suppressed, with a comment naming the rule owner |
Keep the rule set small and reviewed. Every suppression rule is a decision not to look, so record who approved it and when it expires.
Decision guidance
- Small estates (a few accounts): one delegated administrator, the recommended policy, aggregation into one Region, and EventBridge to email or chat is enough. Enable only FSBP to control cost.
- Large estates: custom configuration policies per OU, resource criticality rules driven by tags, ticketing integration with ownership routing, and a SIEM fed from the aggregation Region.
- Allow list or deny list of controls: a deny list keeps new controls enabled automatically (good for coverage); an allow list keeps findings stable (good for regulated reporting). Most organisations should use a deny list.
Framework mapping
| Capability | NIST SP 800-53 Rev. 5 | NIST CSF 2.0 | CIS Controls v8.1 | ISO/IEC 27001:2022 |
|---|---|---|---|---|
| Standards and controls, central configuration | CA-7, CM-6, PM-31 | ID.RA-01, PR.PS-01 | 4 | A.8.9 |
| Inspector and exposure findings | RA-5 | ID.RA-01 | 7 | A.8.8 |
| GuardDuty and correlated findings | SI-4, IR-5 | DE.CM-09, DE.AE-02 | 13 | A.8.16 |
| Automation, ticketing and response | IR-4 | RS.MA-01 | 17 | A.5.26 |
Common misconfigurations
- Enabled in one Region only: Security Hub CSPM processes findings only in Regions where it is enabled, so activity in an unused Region goes unseen. Link all Regions and block unused ones with SCPs.
- AWS Config missing or partial: controls show as no data rather than failed, which reads as a good score.
- Management account as administrator or deployer: it cannot be the delegated administrator, and running tooling from it widens access to the most powerful account.
- Suppression as a tuning strategy: rules that suppress whole controls hide real failures and still incur check charges; disable the control and record the decision instead.
- Local configuration only: auto-enable without central configuration covers new accounts in the current Region only and lets members drift.
- Findings with no owner: without resource criticality, tags and routing, thousands of MEDIUM findings bury the few CRITICAL ones.
Practitioner checklist
- The Security Tooling account is the delegated administrator for Security Hub, Security Hub CSPM, GuardDuty, Inspector, Macie and Detective.
- A non opt-in home Region aggregates every Region in use; unused Regions are denied by SCP.
- AWS Config records in every account and Region, including the delegated administrator.
- Central configuration policies are applied at root and OU level; no unexplained self-managed accounts.
- FSBP is enabled everywhere; other standards are applied by OU where required.
- SCPs block disabling Security Hub, leaving the organisation and stopping AWS Config.
- Analysts use a triage role; configuration and automation rule changes need an elevated role and are alerted on.
- CRITICAL and HIGH findings reach on-call within minutes through EventBridge, Jira Cloud or ServiceNow.
- Suppression rules and disabled controls each have an owner and a review date.
Standards and references
- AWS: What is AWS Security Hub?
- AWS: What is AWS Security Hub CSPM?
- AWS: Central configuration in Security Hub CSPM
- AWS: Standards reference for Security Hub CSPM
- AWS: Automation rules in Security Hub
- AWS Prescriptive Guidance: Security Reference Architecture, Security Tooling account
- AWS: AWS managed policies for Security Hub
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: AWS Security Guide · AWS Config Guide · AWS Control Tower Guide · Prowler and Cloud Posture Scanning Guide · Cloud Detection Engineering Guide