In an Auto Scaling group, instances are ephemeral-local disks and instance-level log files can disappear during scale-in or replacement. The most durable, operationally simple pattern is tostream logs off-host continuouslyto a managed log service. Installing and configuring theCloudWatch agent(or unified agent) to ship application logs toAmazon CloudWatch Logsensures logs are centralized and remain available regardless of instance lifecycle events. This directly solves the "logs lost after scale-in" problem and provides high availability for audit and investigation.
CloudWatch Logs also supports retention controls. The security engineer can set the log group retention toat least 1 year(or longer), meeting the audit requirement without building custom storage workflows. Access can be controlled with IAM to restrict who can view or export logs, and CloudWatch logs can be further integrated with Athena/OpenSearch/SIEM tools if needed.
Option A adds complexity and still ties durability to managing volumes across instance churn, with operational risk and scaling challenges. Option B requires daily copy jobs and can still lose logs between copy intervals; it also adds shared filesystem management overhead. Option D is manual and does not ensure durability, and it introduces operational friction during scale-in. Therefore, centralized log shipping to CloudWatch Logs is the best recommendation.

AWS WAF supports logging of detailed HTTP request information, including source IP addresses, request URIs, headers, and rule evaluation results. According to the AWS Certified Security - Specialty documentation,Amazon S3 combined with Amazon Athenais the recommended and most cost-effective solution for ad hoc and forensic analysis of AWS WAF logs.
By configuring AWS WAF to deliver logs to Amazon S3 and usingAthena with partition projection, the security engineer can efficiently query large volumes of log data without maintaining partitions manually.
This enables rapid identification of application-layer attacks such as SQL injection, cross-site scripting, and bot activity.
Options A and D are incorrect because AWS WAF logs are not delivered to CloudTrail. Option B is invalid because OpenSearch cannot directly query data stored in S3 without ingestion or additional tooling.
AWS documentation highlightsS3 + Athenaas a best practice for scalable, serverless analysis of AWS WAF logs.
* AWS Certified Security - Specialty Official Study Guide
* AWS WAF Logging Documentation
* Amazon Athena Best Practices

AWS KMS key policies can restrict how and when a key is used by applyingconditions such as kms:
ViaService, which limits usage to requests that originate from a specific AWS service. According to the AWS Certified Security - Specialty Official Study Guide and AWS KMS documentation, the kms:ViaService condition is evaluated against the service that calls KMS on behalf of the principal.
Using StringEquals with kms:ViaService restricts usage toexactly one service endpoint. However, AWS services can invoke KMS throughservice variants, internal endpoints, or additional service integrations. When StringEquals is used, these variations can unintentionally bypass the condition, allowing the key to be used by other services through different internal service paths.
Changing the condition operator from StringEquals to StringLike ensures thatonly EC2-related service callsthat match the intended service pattern are allowed, while still preventing use by unrelated AWS services.
This aligns with AWS guidance to use StringLike when service invocation patterns may vary.
Option B is incorrect because the root principal statement is required to retain administrative control over the key. Option C is invalid because changing Regions does not address unauthorized service usage. Option D does not restrict key usage and does not mitigate the issue.
AWS documentation explicitly recommendstightening condition operatorsin KMS key policies to prevent unintended service access while maintaining required functionality.
* AWS Certified Security - Specialty Official Study Guide
* AWS Key Management Service Developer Guide
* AWS KMS Key Policy Best Practices

Option C best meets the requirements because Amazon GuardDuty provides Kubernetes-focused threat detection for Amazon EKS by analyzingEKS control plane audit logs(EKS Protection) and combining that signal withruntime telemetryfrom the worker nodes (Runtime Monitoring). EKS audit logs capture Kubernetes API activity and authorization decisions, allowing GuardDuty to detect suspicious cluster actions such as unusual API calls, unexpected access patterns, or indicators of compromise within the cluster.
Runtime Monitoring extends coverage tooperating system/process activity, network connections, and file activityon the nodes, which directly aligns with the need to monitor OS, networking, and file events in addition to audit logs.
For notifications, GuardDuty generatesfindingsthat can be delivered throughAmazon EventBridgerules.
EventBridge can route relevant GuardDuty findings to anAmazon SNS topic, and SNS can sendemail alertsto the security team by subscribing the team's mailing list to the topic. This approach is fully managed, near real time, and avoids building custom log-parsing pipelines while still providing actionable alerts based on GuardDuty's curated EKS threat detections.

Cross-account access for an external auditor typically usesSTS AssumeRoleinto a role that exists in each target account. Three common failure points are: (1) therole ARNbeing wrong or missing, (2) the auditor lacking permission to callsts:AssumeRoleon that role, and (3) anexternal IDmismatch when the role trust policy requires it. Theexternal IDis a best practice for third-party access to mitigate the confused-deputy problem; if the trust policy includes an sts:ExternalId condition, the auditor must supply the exact expected value. If it's absent or incorrect, AssumeRole will be denied.
The auditor must also be authorized on their side (IAM user/role policy) to call sts:AssumeRole (permission), and the destination role'strust policymust trust the auditor principal. If either side is misconfigured, the assume operation fails. Finally, the auditor must reference the correctrole ARNfor each account; using an incorrect ARN (wrong account ID/role name/path) is a frequent reason only "some accounts" fail.
Incorrect password (B) is irrelevant if the auditor is using API/STS federation rather than console password auth, and EC2 instance roles (D) are not required for an external auditor. Missing/incorrect secret key (E) is possible for API auth, but the question centers on cross-account role access issues; the canonical causes are external ID, AssumeRole permission, and role ARN.