We choose AWS configuration tools around the question we need to answer: what changed, which control failed, or who can access a resource. Those answers often need different evidence.
AWS Config records supported resource configurations and evaluates rules. Security Hub CSPM runs security-control checks and brings findings together. IAM Access Analyzer examines access and policies through several analysis capabilities. They can work together, but each needs an explicit scope and someone responsible for the findings.
Compare the evidence each service produces
| Service | Useful question | Evidence to inspect |
|---|---|---|
| AWS Config | What was this resource’s configuration, and does it meet this rule? | Recorded configuration, evaluation time, rule scope, and result |
| Security Hub CSPM | Which enabled controls or integrated tools report a problem? | Control or provider, resource identifier, finding status, and collection coverage |
| IAM Access Analyzer | What access does this analysis identify, and is it intended? | Analyzer type, trust boundary, policy context, and affected principal or resource |
The AWS Config overview describes configuration history and relationships. Its usefulness depends on what you record. A missing resource type or account can leave a gap that a quiet dashboard will not explain for you.
Security Hub CSPM adds control findings and findings from integrated products. Treat a finding’s origin as part of the evidence. A control result and a threat-detection alert may refer to the same resource while requiring different investigations.
IAM Access Analyzer includes external, internal, and unused-access analysis, along with policy checks and policy generation. Check the capabilities and supported resources for the analyzer you actually enabled. The product name alone does not establish coverage.
Start with a concrete resource
Suppose a hypothetical reporting application keeps private exports in an S3 bucket and shares selected files with a partner account. The intended arrangement is narrow: the partner needs one set of exports, while the application writes new reports.
We would record that intended access before reviewing findings. Otherwise, the team may archive a warning because cross-account sharing sounds normal, or remove a legitimate relationship without understanding the workload.
Configuration history can help establish what changed. A security-control finding can identify a setting that conflicts with an enabled check. An access-analysis result can point to a sharing relationship worth reviewing. None of those observations alone tells you whether the partner downloaded an inappropriate file. That requires the relevant activity evidence and context.
Our broader cloud misconfiguration guide covers the difference between a problematic setting and evidence of misuse. Keep that distinction in the ticket so the next reviewer knows what has actually been observed.
Write a coverage checklist before enabling more controls
For each environment, keep a coverage record with:
- Which accounts and regions are included, and which are intentionally excluded.
- Which resource types are recorded or analyzed.
- Which standards, controls, and integrations are enabled.
- When the most recent successful collection or evaluation occurred.
- Who owns gaps, exceptions, and failed collection jobs.
Check a resource you know should appear. Then make a harmless configuration change in a test environment that should trigger a finding from an enabled check. Follow it through collection, evaluation, notification, and the team’s work queue. Check that an owner receives a finding with enough context to act on it.
Avoid assuming that central administration automatically covers every new account or region. Review the deployed configuration and the onboarding process. Product capabilities and organization settings can change independently.
Decide what happens after a finding arrives
We would route findings by the action needed. A missing log destination may belong to the platform team. Unexpected partner access needs the data owner and identity owner. Evidence of unauthorized activity should enter the incident-response process.
Include the resource, observed condition, collection time, expected condition, and a safe way to check the repair. For the reporting bucket, a useful retest would exercise both the partner’s permitted read and a read outside its intended scope using harmless objects.
Keep duplicate findings connected. Closing the security-control ticket should not silently close an investigation into actual access. Conversely, three tools reporting one setting should not create three independent teams making conflicting changes.
Separate evaluation from automatic repair
AWS Config’s evaluation guidance explicitly distinguishes proactive evaluation from enforcement: a failed proactive check does not itself block deployment or fix the resource.
Before adding automated remediation, document the permitted change, its owner, exceptions, and how to recover if it disrupts a legitimate workflow. Start in a test environment with representative cases. A policy that closes public access may be appropriate for private reports and inappropriate for an intentionally public publishing bucket.
We would choose the services around these questions and test the complete response path. For help mapping the findings to your environment, our cloud services cover infrastructure and configuration work. Bring the accounts in scope and a few representative findings so we can trace them to the work your team needs to do.
