Join our Newsletter — 33% off our NHI Course
Home› Guides› AWS Config Guide: Rules, Conformance Packs and Configuration…
Guide Cloud Security

AWS Config Guide: Rules, Conformance Packs and Configuration Recorder

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 8 November 2026 13 min read
On this page

AWS Config is the record of what every AWS resource looked like and when it changed, and when it is switched off, scoped badly or left to drift, both drift detection and audit evidence across the estate quietly disappear. Security Hub controls, Firewall Manager, conformance packs and many compliance tools depend on it, so its configuration is a security control in its own right. This guide covers how the configuration recorder, delivery channel, rules, conformance packs, remediation and aggregators work, who should administer them, how to protect Config from being disabled, the cost and scope decisions that trip teams up, and a hardening baseline with rollout steps. It is written for cloud security engineers, platform teams and auditors working in multi-account AWS environments. For the wider AWS security service set, 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

  • The recorder is the foundation: one customer managed configuration recorder per account per Region, and by default it does not record the global IAM resource types, so IAM changes are invisible until you choose to record them in one Region.
  • Rules detect, they do not prevent: detective rules evaluate deployed resources and proactive rules evaluate properties before deployment, but AWS states that proactive rules neither block deployment nor remediate.
  • Conformance packs package rules for a framework, but AWS is explicit that the sample packs do not ensure compliance or guarantee an assessment pass.
  • Automatic remediation needs care: it acts on a periodic compliance snapshot and AWS warns it can run against resources that are already compliant.
  • Protect Config itself: deny StopConfigurationRecorder and delivery channel deletion with service control policies, and alert on recorder state changes.

How identity and access work in AWS Config

AWS Config has three identity layers, and confusing them is a common source of gaps.

IdentityWhat it doesHardening point
Recorder roleLets Config read resource configurations and write to the delivery channel. The service-linked role (AWSServiceRoleForConfig) is predefined, and its policy cannot be attached to other entities; a custom role using the AWS managed AWS_ConfigRole policy is the alternativePrefer the service-linked role. IAM and Organizations policies can prevent Config from recording changes, and rules do not account for that, so test SCPs against the recorder
Remediation roleThe role an SSM Automation document assumes (passed as AutomationAssumeRole) to change non-compliant resourcesScope it per remediation, never a single admin role for all remediations; restrict iam:PassRole
Human and pipeline administratorsCreate rules, conformance packs, remediation and aggregatorsSeparate who can read compliance data from who can change rules or stop the recorder

Service-linked recorders are a newer layer: AWS services such as Security Hub, Security Hub CSPM and CloudWatch can create their own recorders (for example AWSConfigurationRecorderForSecurityHubCSPM) whose scope is set by the linked service and cannot be changed directly. A service-linked recorder with INTERNAL scope is free and delivers no configuration items to your delivery channel; one with PAID scope delivers them and is charged. Know which recorders exist in each account before you reason about coverage.

Admin roles and privileged access

  • Delegated administration: in AWS Organizations, deploy organisation rules and organisation conformance packs from a delegated administrator account in the security tooling account, not from the management account.
  • Stop and delete rights: config:StopConfigurationRecorder, config:DeleteConfigurationRecorder, config:DeleteDeliveryChannel and changes to the recorder's resource scope are the permissions an attacker wants. Deny them in an SCP for everyone except a named platform role. See the AWS Organizations and SCPs Guide.
  • Rule tampering: deleting or editing a rule, or adding an exclusion to a conformance pack, removes detection without stopping the recorder. Treat rule administration as privileged and manage rules through infrastructure as code with review.
  • Remediation power: a remediation role that can modify security groups, bucket policies or IAM is effectively an administrator. Review its permissions with the same care as a human admin role.
  • Evidence integrity: limit who can write to or delete from the delivery channel bucket. Keep S3 Object Lock default retention off on the delivery bucket, as AWS Config's delivery requirements specify, and protect history with bucket policies, versioning and replication to the log archive account instead.

Integrations, API tokens and service accounts

Config is an integration hub, and the configuration items it records are also a detailed map of your estate, so the data deserves protection.

  • Delivery channel: one per account per Region, required to use Config. History and snapshot files go to S3, encrypted with SSE-S3 by default or with a KMS key you supply; change notifications go to an SNS topic you choose. Centralise delivery into the log archive account and use a customer managed key with a restrictive key policy.
  • Aggregators: collect configuration and compliance data across accounts and Regions into one account. They are read-only: you cannot deploy rules or push data through an aggregator. Organisation sources need no per-account authorisation; individual account sources must each authorise the aggregator account and Region. There is no extra charge.
  • Security Hub and Firewall Manager: Security Hub controls run as service-linked Config rules, and Firewall Manager depends on continuous recording, so switching an account to daily recording breaks Firewall Manager policies. See the AWS Security Hub Guide.
  • Custom Lambda rules: each function runs with its own execution role and should be scoped to specific resource types; AWS warns that a custom Lambda rule without a resource scope invokes the function for every resource in the account.
  • Third-party tools: posture management and compliance platforms often read Config through cross-account roles. Grant them read-only access to the aggregator account rather than to every member account. See the Compliance Automation Guide.

Logging and detection

Config is itself a detection source, and it must be monitored so its silence is noticed.

How rules evaluate

SettingOptionsNotes
Rule typeManaged, custom Lambda, custom policy (Guard), service-linked, organisationalManaged rules come with parameters; default values are pre-populated only in the console, not the API or CLI
TriggerConfiguration changes, periodic, hybridChange triggers can be scoped by resource type, resource ID or tag
Evaluation modeDetective, proactiveProactive evaluates proposed properties; it does not block or remediate
ResultCOMPLIANT, NON_COMPLIANT, ERROR, NOT_APPLICABLEERROR means a parameter problem, which silently hides real results if nobody watches it
Recording frequencyContinuous, dailyDaily produces at most one configuration item per resource per 24 hours; Firewall Manager needs continuous

What to alert on

EventSourceATT&CK v20
Recorder stopped, deleted or scope reducedCloudTrail (StopConfigurationRecorder, DeleteConfigurationRecorder, PutConfigurationRecorder); the AWS::Config::ConfigurationRecorder configuration item records recorder state at no chargeT1685 Disable or Modify Tools
Delivery channel deleted or redirectedCloudTrail DeleteDeliveryChannel, PutDeliveryChannelT1685 Disable or Modify Tools
Security group opened to the internetConfig rule on security group ingress; configuration item changeT1578 Modify Cloud Compute Infrastructure
IAM policy or role change outside pipelinesConfiguration items for IAM types (if recorded)T1098 Account Manipulation
Access keys older than policyaccess-keys-rotated rule with maxAccessKeyAgeT1078.004 Cloud Accounts

Route compliance change events through EventBridge into your SIEM or ticketing, and correlate configuration items with CloudTrail to answer who made a change, since Config records what changed and CloudTrail records the call. See the AWS CloudTrail Guide.

Metrics: percentage of account and Region pairs with a running recorder and delivery channel (target 100% of enabled Regions); IAM global types recorded in exactly one Region; rules in ERROR state; mean time from NON_COMPLIANT to fixed for high-severity rules; conformance pack compliance score trend; automatic remediation failures per week.

Hardening baseline

Scope and cost decisions

  • Recording strategy: use "record all current and future resource types with exclusions" (EXCLUSION_BY_RESOURCE_TYPES) rather than an inclusion list, so new resource types are recorded automatically. Use inclusion (INCLUSION_BY_RESOURCE_TYPES) only for accounts with a narrow, stable purpose.
  • Global IAM types: record AWS::IAM::User, Role, Group and Policy once, in one supported Region, to avoid duplicate items and throttling. They cannot be recorded in several newer Regions, and periodic IAM rules such as iam-user-mfa-enabled and root-account-mfa-enabled should be deployed only where they make sense.
  • Home-Region-only types: some global resources are recorded only in a home Region, for example CloudFront distributions, Route 53 hosted zones and WAFv2 web ACLs in US East (N. Virginia). Make sure that Region is enabled for recording even if you run no workloads there.
  • Ephemeral workloads: Spot Instances, EMR and Auto Scaling create large numbers of configuration items. Exclude those types, or run such workloads in separate accounts, rather than switching the whole account to daily recording.
  • Continuous versus daily: keep continuous recording for security-relevant types (IAM, security groups, S3, KMS, network); daily recording is a cost tool for low-risk, high-churn types. AWS notes the first month costs more because Config evaluates every selected resource during bootstrapping.

Conformance packs and rules

A conformance pack is a YAML template of Config rules and remediation actions deployed as one unit in an account and Region, or across an organisation. AWS publishes sample templates, including Operational Best Practices for NIST 800-53 rev 5, NIST CSF, CIS AWS Foundations Benchmark v1.4 Level 1 and Level 2, and PCI DSS 4.0, plus Security Best Practices packs for services such as Amazon EKS, AWS Lambda, Amazon RDS and AWS Secrets Manager. AWS states that these templates "are not designed to ensure your compliance with a specific governance standard. They can neither replace your internal efforts nor guarantee that you will pass a compliance assessment." Use them as a starting point, remove rules that duplicate Security Hub controls you already run, and add rules for your own policies.

Decision guidance: a small organisation should rely on Security Hub's standards for most posture checks and add a handful of custom Config rules for local policy, because running a full NIST pack alongside Security Hub doubles evaluations and cost. A large or regulated organisation should deploy an organisation conformance pack per framework it must evidence, from a delegated administrator, and use the aggregator as the single evidence source for auditors. See the Audit Evidence and Continuous Controls Monitoring Guide.

Remediation

Remediation runs AWS Systems Manager Automation documents, either AWS managed or your own, manually or automatically. Retries can be configured, for example 5 retries in 300 seconds, and each run has a cost. AWS warns that automatic remediation "relies on the compliance data snapshot which is captured on a periodic basis" and "can be initiated even for compliant resources", and remediation can be managed only for rules that are not service-linked. Start with manual remediation, move a rule to automatic only after the document has been tested in a non-production account, and choose low-risk, idempotent actions first, such as enabling bucket encryption or versioning, rather than deleting security group rules in production. See the AWS Systems Manager Security Guide.

Phased rollout

  1. Inventory recorders: in every account and enabled Region, list customer managed and service-linked recorders, their scope and frequency, and the delivery channel target.
  2. Standardise recording: deploy the recorder and delivery channel through Control Tower or StackSets with exclusion-based scope, IAM types in one Region and delivery to the log archive account. See the AWS Control Tower Guide.
  3. Protect Config: apply the SCP denying recorder and delivery channel changes, test it in a sandbox OU first, and add CloudTrail alerts.
  4. Aggregate: create an organisation aggregator in the security tooling account.
  5. Rules in observe mode: deploy conformance packs without remediation and review results for two to four weeks, tuning parameters and fixing ERROR rules.
  6. Proactive checks in pipelines: use proactive evaluation for supported resource types in CloudFormation pre-deployment checks, remembering that it informs and does not block on its own.
  7. Remediate gradually: enable automatic remediation rule by rule, starting in non-production. Roll back by removing the remediation configuration from the rule; the rule itself keeps detecting.

Framework mapping

Config capabilityNIST SP 800-53 Rev. 5NIST CSF 2.0CIS Controls v8.1ISO/IEC 27001:2022
Recorder and resource inventoryCM-8ID.AM-021, 2A.5.9
Rules and conformance packsCM-2, CM-6, CA-7PR.PS-014A.8.9
Change historyCM-3, AU-12PR.PS-04, DE.CM-098A.8.15, A.8.32
Protecting delivery dataAU-9, AU-11PR.DS-018A.8.15
RemediationSI-2, CM-6ID.IM-014, 7A.8.8, A.8.9

Common misconfigurations

  • Recording only in "our" Regions: an attacker launches resources in an unused Region where no recorder runs, and nothing records them. Either record in every enabled Region or disable unused Regions with an SCP.
  • IAM types never recorded: the default excludes them, so role and policy changes, the most important identity drift, never appear in Config history.
  • Daily recording everywhere to cut cost: Firewall Manager policies stop working and a security group can be opened and closed within a day without a trace in Config.
  • Treating a green conformance pack as compliance: the pack is presented to auditors as proof of PCI DSS or NIST compliance, against AWS's own disclaimer, and controls outside the pack's rules are never evidenced.
  • Over-privileged remediation: one shared remediation role with administrator rights, which becomes a privilege escalation path for anyone who can attach a remediation document to a rule.
  • Nobody watching ERROR results or empty scope: rules with bad parameters or out-of-scope resources look quiet and are read as compliant.

Practitioner checklist

  • Every account and enabled Region has a running recorder and delivery channel, deployed as code.
  • Recording uses EXCLUSION_BY_RESOURCE_TYPES, with IAM global types recorded in exactly one Region and the home Region for global services enabled.
  • Security-relevant resource types use continuous recording; daily recording is limited to documented low-risk types.
  • Delivery goes to the log archive account, encrypted with a customer managed key, with versioning and restricted delete rights.
  • An SCP denies stopping or deleting the recorder and delivery channel, and CloudTrail alerts fire on any attempt.
  • An organisation aggregator in the security tooling account is the single source for compliance reporting.
  • Conformance packs are deployed from a delegated administrator, tuned, and not presented as certification.
  • Automatic remediation is enabled only for tested, idempotent documents, each with a narrowly scoped role.
  • Rules in ERROR state and remediation failures are reviewed weekly.

Standards and references

Related NHI Mgmt Group resources: AWS Security Guide · Configuration Drift and Baseline Enforcement Guide · AWS Security Hub Guide · Cloud Misconfiguration Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

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 8 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

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.