The AWS Well-Architected Security Pillar is the review framework most AWS workloads are measured against, yet many reviews end as a questionnaire filled in from memory, with no evidence and no fixes. The pillar asks eleven questions, SEC 1 to SEC 11, backed by 63 named best practices, and used well it gives architects, security teams and AWS partners a shared language for what "secure enough" means for a workload. This guide explains the structure of the pillar, what each question really tests, how to gather evidence for the answers from AWS services rather than opinion, how to run a review that produces a prioritised improvement plan, and how the pillar maps to NIST, CIS and ISO frameworks. It is written for cloud architects, security reviewers and partner teams who run or receive Well-Architected reviews. For the AWS security services themselves, 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
- Seven areas, eleven questions, 63 best practices: the current pillar (published 6 November 2024) organises security into foundations, identity and access, detection, infrastructure, data, incident response and application security.
- Answer with evidence, not belief: almost every best practice can be checked against AWS Config, IAM Access Analyzer, Security Hub, CloudTrail or Organizations data before the review meeting.
- SEC 1 to SEC 3 carry the most risk: account separation, root user protection, temporary credentials and least privilege decide the blast radius of everything else.
- A review is only useful if it ends in owned, dated fixes: record high risk issues, assign each to an owner, and re-review at a milestone rather than starting again next year.
- The pillar is a review framework, not a compliance standard: map it to NIST CSF 2.0, SP 800-53 or ISO/IEC 27001 when you need audit evidence, and do not present a clean review as certification.
What the Security Pillar is and how it is structured
The AWS Well-Architected Framework has six pillars: Operational Excellence, Security, Reliability, Performance Efficiency, Cost Optimization and Sustainability. The Security Pillar whitepaper is versioned by date; the current edition is dated 6 November 2024, which updated guidance in SEC 2, 3, 4, 6, 7, 8, 9, 10 and 11, following a large reorganisation in June 2024 that reordered and consolidated best practices. Check the document revisions page before a review, because best practice numbers have moved between editions.
The pillar states seven design principles:
- Implement a strong identity foundation
- Maintain traceability
- Apply security at all layers
- Automate security best practices
- Protect data in transit and at rest
- Keep people away from data
- Prepare for security events
Each question in a review belongs to one of seven best practice areas, and each question lists its best practices. A reviewer selects which best practices the workload has implemented; questions with important best practices missing are flagged as risks.
| Area | Question | Best practices |
|---|---|---|
| Security foundations | SEC 1. How do you securely operate your workload? | SEC01-BP01 to BP08 |
| Identity and access management | SEC 2. How do you manage authentication for people and machines? | SEC02-BP01 to BP06 |
| Identity and access management | SEC 3. How do you manage permissions for people and machines? | SEC03-BP01 to BP09 |
| Detection | SEC 4. How do you detect and investigate security events? | SEC04-BP01 to BP04 |
| Infrastructure protection | SEC 5. How do you protect your network resources? | SEC05-BP01 to BP04 |
| Infrastructure protection | SEC 6. How do you protect your compute resources? | SEC06-BP01 to BP05 |
| Data protection | SEC 7. How do you classify your data? | SEC07-BP01 to BP04 |
| Data protection | SEC 8. How do you protect your data at rest? | SEC08-BP01 to BP04 |
| Data protection | SEC 9. How do you protect your data in transit? | SEC09-BP01 to BP03 |
| Incident response | SEC 10. How do you anticipate, respond to, and recover from incidents? | SEC10-BP01 to BP08 |
| Application security | SEC 11. How do you incorporate and validate the security properties of applications throughout the design, development, and deployment lifecycle? | SEC11-BP01 to BP08 |
The questions in practice: what each one really tests
SEC 1: security foundations
The eight best practices are: SEC01-BP01 Separate workloads using accounts; BP02 Secure account root user and properties; BP03 Identify and validate control objectives; BP04 Stay up to date with security threats and recommendations; BP05 Reduce security management scope; BP06 Automate deployment of standard security controls; BP07 Identify threats and prioritize mitigations using a threat model; BP08 Evaluate and implement new security services and features regularly. The two that most often fail are BP01 (production and development sharing one account) and BP07 (no threat model at all). See the AWS Multi-Account Security Reference Architecture Guide and the AWS Control Tower Guide.
SEC 2 and SEC 3: identity and permissions
SEC 2 covers authentication: BP01 Use strong sign-in mechanisms, BP02 Use temporary credentials, BP03 Store and use secrets securely, BP04 Rely on a centralized identity provider, BP05 Audit and rotate credentials periodically, BP06 Employ user groups and attributes. SEC 3 covers permissions: BP01 Define access requirements, BP02 Grant least privilege access, BP03 Establish emergency access process, BP04 Reduce permissions continuously, BP05 Define permission guardrails for your organization, BP06 Manage access based on lifecycle, BP07 Analyze public and cross-account access, BP08 Share resources securely within your organization, BP09 Share resources securely with a third party. Machine identities are named explicitly in both questions, which is where reviews most often find long-lived IAM user access keys used by pipelines and integrations. See the AWS Access Keys Security Guide and the AWS IAM Policies and Roles Guide.
SEC 4: detection
BP01 Configure service and application logging; BP02 Capture logs, findings, and metrics in standardized locations; BP03 Correlate and enrich security alerts; BP04 Initiate remediation for non-compliant resources. A workload passes BP01 only if application logs exist as well as CloudTrail; BP04 expects automated remediation, not a ticket queue. See the AWS CloudTrail Guide and the AWS Config Guide.
SEC 5 and SEC 6: infrastructure protection
SEC 5: BP01 Create network layers, BP02 Control traffic flow within your network layers, BP03 Implement inspection-based protection, BP04 Automate network protection. SEC 6: BP01 Perform vulnerability management, BP02 Provision compute from hardened images, BP03 Reduce manual management and interactive access, BP04 Validate software integrity, BP05 Automate compute protection. SEC06-BP03 is the practical test for SSH keys and bastion hosts: interactive access should go through managed session tooling with logging. See the Amazon VPC Security Guide and the EC2 Instance Metadata Service Hardening Guide.
SEC 7 to SEC 9: data protection
SEC 7: BP01 Understand your data classification scheme, BP02 Apply data protection controls based on data sensitivity, BP03 Automate identification and classification, BP04 Define scalable data lifecycle management. SEC 8: BP01 Implement secure key management, BP02 Enforce encryption at rest, BP03 Automate data at rest protection, BP04 Enforce access control. SEC 9: BP01 Implement secure key and certificate management, BP02 Enforce encryption in transit, BP03 Authenticate network communications. The word that matters is "enforce": a review should look for preventive controls (default encryption, bucket policies denying non-TLS requests, SCPs) rather than a statement that teams turn encryption on. See the Cloud Customer-Managed Keys and BYOK/HYOK Decision Guide.
SEC 10: incident response
BP01 Identify key personnel and external resources; BP02 Develop incident management plans; BP03 Prepare forensic capabilities; BP04 Develop and test security incident response playbooks; BP05 Pre-provision access; BP06 Pre-deploy tools; BP07 Run simulations; BP08 Establish a framework for learning from incidents (added in October 2023). BP05 and BP06 are the cloud-specific ones: responders need break-glass roles and forensic tooling in place before an incident, not requested during one. See the Cloud Incident Response Guide.
SEC 11: application security
BP01 Train for application security; BP02 Automate testing throughout the development and release lifecycle; BP03 Perform regular penetration testing; BP04 Conduct code reviews; BP05 Centralize services for packages and dependencies; BP06 Deploy software programmatically; BP07 Regularly assess security properties of the pipelines; BP08 Build a program that embeds security ownership in workload teams. BP07 brings CI/CD pipelines and their credentials into scope, which many reviews skip. See the DevSecOps Guide.
Gathering evidence before the review
The most common weakness of a Well-Architected review is that answers come from the people who built the workload, in a meeting, with nothing checked. Collect evidence beforehand so the meeting is spent on gaps and decisions. The mapping below is a practical starting point; the service choices are ours, not part of the pillar.
| Best practice | Evidence to collect | Typical source |
|---|---|---|
| SEC01-BP01 Separate workloads using accounts | Account list per environment; OU structure | AWS Organizations |
| SEC01-BP02 Secure account root user | Root MFA status, root access keys absent, recent root sign-ins | IAM credential report, CloudTrail |
| SEC02-BP02 Use temporary credentials | IAM users with active access keys and their last-used dates | IAM credential report |
| SEC03-BP04 Reduce permissions continuously | Unused roles, unused permissions | IAM Access Analyzer, last-accessed data |
| SEC03-BP07 Analyze public and cross-account access | External access findings for S3, KMS, IAM roles and other resources | IAM Access Analyzer |
| SEC04-BP01/BP02 Logging in standard locations | Organisation trail, log archive account, application log destinations | CloudTrail, CloudWatch Logs |
| SEC04-BP04 Remediate non-compliant resources | Config rules, conformance pack compliance, remediation actions | AWS Config, Security Hub |
| SEC05-BP02 Control traffic flow | Security groups open to 0.0.0.0/0, VPC flow logs status | Config rules, VPC Flow Logs |
| SEC06-BP01 Vulnerability management | Open critical findings and age | Amazon Inspector |
| SEC07-BP03 Automate classification | Sensitive data discovery results for S3 | Amazon Macie |
| SEC08-BP02 Enforce encryption at rest | Default encryption settings, unencrypted volumes and databases | Config rules, Security Hub controls |
| SEC10-BP07 Run simulations | Date and output of the last game day or tabletop | Incident response records |
The AWS Well-Architected Tool supports this approach: it holds the questions as lenses, lets you add custom lenses with your organisation's own questions, and integrates with AWS Trusted Advisor and AWS Service Catalog AppRegistry to help find the information needed to answer questions. For automated posture evidence across many accounts, see the Prowler and Cloud Posture Scanning Guide.
Running the review: a step-by-step approach
- Scope the workload: define the accounts, Regions, data classes and dependencies in scope. A review of "the platform" with no boundary produces generic answers.
- Collect evidence two weeks ahead: run the evidence queries above and share the output with the workload team so surprises are found before the meeting.
- Run the questions with the right people: the workload owner, an engineer who operates it, and an independent security reviewer. Answer each best practice as implemented, partly implemented (record as not implemented, with a note) or not applicable with a reason.
- Prioritise risks: rank by exposure, not by question order. A missing SEC02-BP02 on a pipeline with production write access outranks a missing SEC07-BP04.
- Build the improvement plan: each risk gets an owner, a target date and an expected evidence change, for example "IAM users with access keys: 14 to 0".
- Save a baseline and re-review: record the review state, then re-review after the planned fixes or at a fixed interval, comparing against the baseline rather than starting fresh.
- Handle exceptions properly: where a risk is accepted, record who accepted it, why and until when, so the next review does not rediscover it as new.
Decision guidance: a small organisation with a few workloads should review its most sensitive workload first and treat the result as its AWS security baseline; a full review of every workload will stall. A large organisation should make SEC 1 to SEC 3 a platform-level review answered once for the landing zone, then run workload reviews that focus on SEC 4 to SEC 11 and inherit the platform answers. When an AWS partner runs the review, ask for the evidence behind each answer and agree in advance whether the output is for internal improvement or for a funding or compliance purpose, because the incentives differ.
Detection and measurement
A review is a point in time. Turn its findings into signals that show whether the workload stays at the reviewed level.
- Risk count trend: high risk items open per workload, and the change between reviews.
- Improvement plan velocity: percentage of review actions closed by their target date.
- Credential posture: count of long-lived access keys and their age (SEC 2), unused roles older than 90 days (SEC 3).
- Control drift: Config rule non-compliance for controls the review relied on, such as default encryption and security group exposure (SEC 4, 5, 8).
- Response readiness: days since the last incident simulation and whether pre-provisioned responder access was tested (SEC 10).
- Coverage: percentage of production workloads reviewed in the last 12 months.
Framework mapping
The pillar does not publish an official mapping to other frameworks. The mapping below is indicative, area by area, for teams that need to reuse review evidence for audit.
| Pillar area | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 | CIS Controls v8.1 | ISO/IEC 27001:2022 Annex A |
|---|---|---|---|---|
| Security foundations (SEC 1) | GV.RM-01, GV.PO-01 | PM-9, RA-3, SA-8 | 1, 4 | A.5.1, A.5.23 |
| Identity and access (SEC 2, 3) | PR.AA-01, PR.AA-05 | AC-6, IA-2, IA-5 | 5, 6 | A.5.15, A.5.16, A.8.2 |
| Detection (SEC 4) | DE.CM-09 | AU-2, AU-6, CA-7 | 8, 13 | A.8.15, A.8.16 |
| Infrastructure protection (SEC 5, 6) | PR.IR-01, PR.PS-01 | SC-7, CM-6, SI-2 | 4, 7, 12 | A.8.8, A.8.9, A.8.20 |
| Data protection (SEC 7, 8, 9) | PR.DS-01, PR.DS-02 | SC-12, SC-13, SC-28, SC-8 | 3 | A.5.12, A.8.24 |
| Incident response (SEC 10) | RS.MA-01, ID.IM-02 | IR-3, IR-4, IR-8 | 17 | A.5.24, A.5.26, A.5.27 |
| Application security (SEC 11) | PR.PS-06 | SA-11 | 16 | A.8.25, A.8.28, A.8.29 |
Common failure patterns
- Self-attested answers: the workload team ticks "use temporary credentials" while the credential report shows active IAM user keys from three years ago.
- Reviewing against an old edition: reviewers use a spreadsheet of best practices from before the June 2024 reorganisation, so IDs and scope no longer match the current pillar or the Well-Architected Tool.
- Everything marked not applicable: SEC 10 and SEC 11 are skipped as "handled by another team", leaving incident readiness and pipeline security unreviewed.
- No owner for findings: the review produces a list of risks and a report, and nobody is assigned to close them, so the next review finds the same issues.
- Treating a review as compliance: a clean Well-Architected review is presented to a customer or auditor as proof of compliance. It is a best practice review, not an attestation.
- Ignoring machine identities: SEC 2 and SEC 3 are answered for people only, missing the workload roles that cause the most damage when abused.
Lessons from incidents
A common cloud attack path shows why evidence-based reviews matter. Server-side request forgery against an application can obtain a role\'s credentials from the instance metadata service (ATT&CK T1552.005 Cloud Instance Metadata API), and a role that can list and read far more S3 buckets than it needs turns that into data theft (T1530 Data from Cloud Storage). In pillar terms the controls are SEC03-BP02 Grant least privilege access and SEC03-BP04 Reduce permissions continuously, SEC06-BP02 Provision compute from hardened images (IMDSv2 and metadata limits), and SEC04-BP03 Correlate and enrich security alerts for unusual machine identity data access. An evidence-based review surfaces over-broad S3 permissions from IAM Access Analyzer and last-accessed data before anyone can abuse them.
Practitioner checklist
- Confirm the current Security Pillar edition and best practice IDs on the document revisions page before each review cycle.
- Define the workload boundary: accounts, Regions, data classes and dependencies.
- Collect evidence for SEC 1 to SEC 4 from Organizations, the IAM credential report, IAM Access Analyzer, CloudTrail and Config before the meeting.
- Answer SEC 2 and SEC 3 for machine identities as well as people, including CI/CD and third-party integrations.
- Do not accept "not applicable" for SEC 10 or SEC 11 without a named team and evidence.
- Give every high risk item an owner, a target date and a measurable evidence change.
- Answer SEC 1 to SEC 3 once at platform level in large estates, and inherit those answers in workload reviews.
- Track control drift between reviews with Config rules and Security Hub controls tied to the review's assumptions.
- Re-review against a saved baseline after fixes, and at least annually for production workloads.
Standards and references
- AWS: Security Pillar, AWS Well-Architected Framework
- AWS: Security Pillar design principles and best practice areas
- AWS: SEC 1. How do you securely operate your workload?
- AWS: SEC 3. How do you manage permissions for people and machines?
- AWS: Security Pillar document revisions
- AWS: What is the AWS Well-Architected Tool?
- NIST: Cybersecurity Framework 2.0
- MITRE ATT&CK: Cloud Instance Metadata API
Related NHI Mgmt Group resources: AWS Security Guide · AWS Multi-Account Security Reference Architecture Guide · Top 10 AWS IAM Mistakes · Cloud Security Architecture Review Template