SOA-C03 domain - 16% of the exam

Security and Compliance

Security and Compliance is 16% of the SOA-C03 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

59
Practice questions
5
Objectives
47
Medium
12
Hard

Where people go wrong

  • Using a per-principal or detective control as an organisation-wide guardrail

    The wrong answers reach for a permissions boundary, an identity-based policy, a Config rule or a conformance pack when the requirement is a restriction that holds across every member account, including the root user. Boundaries and identity policies attach to individual principals and do not bind the root user, and Config rules and conformance packs detect, or at most remediate, after the resource exists. A service control policy with an explicit Deny sets the ceiling for every principal in the targeted accounts.

    See it in a sample question
  • Assigning a finding to the wrong detection service

    The wrong answers name a security service that sounds relevant but looks at different data. GuardDuty analyses log activity for threats, so it cannot classify S3 contents or list package vulnerabilities. AWS Config evaluates resource configuration and cannot read objects. Macie discovers sensitive data in S3 and does not scan compute. Amazon Inspector scans EC2 instances and ECR images for known software vulnerabilities, and each service answers only its own question.

    See it in a sample question
  • Assuming every certificate is renewed by ACM

    The wrong answers obtain the certificate somewhere other than ACM: a third-party authority, a self-signed file, or a copy installed on each instance. ACM renews automatically only the certificates it issued itself. An imported certificate must be reimported manually before expiry, a cron job on each instance is still manual handling, a self-signed certificate is not publicly trusted, and Secrets Manager does not issue or renew TLS certificates. A public certificate requested from ACM and attached to the load balancer listener renews with no manual handling.

    See it in a sample question
  • Confusing what Access Analyzer, the policy simulator and CloudTrail answer

    The wrong answers swap three tools that answer different questions. CloudTrail records API calls that already happened, so it cannot test a policy that is not yet attached. The policy simulator evaluates a chosen action against a policy, but it is not a continuous scanner and has no view of real usage. IAM Access Analyzer reasons over resource and trust policies for external access, and its unused access analyser reports permissions that went unused over its tracking period.

    See it in a sample question
  • Fixing something other than what Trusted Advisor flagged

    The wrong answers apply a sound security control that leaves the flagged condition in place. Default encryption, CloudFront, a subnet ACL, a private subnet, GuardDuty or a password policy each improve security elsewhere. Trusted Advisor evaluates the resource configuration itself, so the check clears only when the open bucket policy, the unrestricted security group rule, the missing trail or the root user without MFA is changed directly.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleSecurity and Compliancehard

An organisation must guarantee that no principal in any member account, including the account root user and any administrator, can leave a designated AWS Region or disable AWS CloudTrail, regardless of the IAM policies those accounts attach to their own users. The accounts are managed in AWS Organizations with all features enabled. Which control enforces this restriction across the member accounts?

  • AAttach an IAM permissions boundary to every user and role in each member account so the boundary caps what those identities are permitted to do across all services.
  • BApply a service control policy through AWS Organizations that denies the disallowed Region and the trail-disabling actions to the organisational units containing the accounts. Correct
  • CCreate an AWS Config conformance pack in each member account that flags any action taken outside the approved Region or any change that disables the trail.
  • DGrant the central security team an IAM role in each account and rely on it to review and revoke any out-of-Region activity or trail changes that appear in the logs.
Use an organisation service control policy to set a permission ceiling that even the member account root user cannot exceed. A service control policy applied through AWS Organizations defines the maximum available permissions for every principal in the targeted accounts and is evaluated before account-level IAM policies, so an explicit deny in an SCP overrides any allow an account grants itself and applies even to the root user. Permissions boundaries are per-identity and removable, while Config and a review role only detect or remediate after the fact, so none of them enforces the hard guardrail an SCP provides.

Why A is wrong: Permissions boundaries must be applied to each identity individually and can be removed by an account administrator, so they cannot guarantee an account-wide limit that even the root user cannot bypass.

Why B is correct: An SCP sets the maximum permissions for every principal in the targeted accounts, including the root user, so a deny for those actions cannot be overridden by any IAM policy the account attaches.

Why C is wrong: A conformance pack detects and reports non-compliant configurations after they occur but does not block the action, so administrators could still operate outside the Region or disable the trail.

Why D is wrong: A cross-account review role enables after-the-fact remediation but depends on detection and manual action, so it cannot prevent the prohibited actions from being performed in the first place.

Free sampleSecurity and Compliancemedium

A compliance programme requires the organisation to discover where personally identifiable information such as credit card numbers and passport details is stored across hundreds of Amazon S3 buckets, and to receive a report classifying which buckets contain sensitive data. The team wants a managed service that performs this discovery rather than building custom scanners. Which service satisfies this data classification requirement?

  • AAmazon GuardDuty, which inspects the contents of objects in the S3 buckets and produces a classification report listing where credit card and passport data is stored.
  • BAWS Config, using a managed rule that reads each object and tags any bucket that holds credit card or passport information as containing sensitive data.
  • CAmazon Macie, which uses managed and custom data identifiers to scan S3 objects and reports the buckets that contain sensitive personally identifiable information. Correct
  • DAmazon Inspector, by enabling its S3 data scanning feature to read each object and report the buckets that contain credit card numbers and passport details.
Choose Amazon Macie to discover and classify sensitive personally identifiable information stored in Amazon S3 across many buckets. Amazon Macie uses managed data identifiers, and optionally custom ones, to sample and inspect object contents in S3 and classify findings by sensitivity, which is precisely the data classification job described. GuardDuty works from logs and access patterns, AWS Config evaluates configuration metadata without reading objects, and Inspector targets vulnerabilities on compute resources, so none of them perform content-level sensitive data discovery in S3.

Why A is wrong: GuardDuty detects threats from log activity and S3 access patterns but does not inspect object contents to classify personally identifiable information, so it cannot build the sensitivity report required.

Why B is wrong: AWS Config evaluates resource configuration metadata and never reads object contents, so it cannot identify which buckets contain personally identifiable information the way the requirement needs.

Why C is correct: Macie is purpose-built to discover and classify sensitive data in S3 using managed identifiers for items like credit card and passport numbers, so it produces exactly the sensitivity report the programme requires.

Why D is wrong: Inspector assesses software vulnerabilities and network exposure on compute and container resources, not the contents of S3 objects, so it does not perform sensitive data discovery.

Free sampleSecurity and Compliancemedium

A team runs a public website on Amazon EC2 instances behind an Application Load Balancer and needs a publicly trusted TLS certificate for its custom domain that AWS renews automatically before expiry, with no manual certificate handling. Where should the certificate be provisioned and attached to satisfy the automatic renewal requirement most directly?

  • ARequest a public certificate in AWS Certificate Manager and attach it to the HTTPS listener on the Application Load Balancer, where ACM renews it automatically. Correct
  • BBuy a certificate from a third-party authority, install it on each EC2 instance, and configure a cron job on every instance to renew the file before it expires.
  • CImport an externally issued certificate into AWS Certificate Manager and attach it to the load balancer listener, relying on ACM to renew the imported material.
  • DStore a self-signed certificate in AWS Secrets Manager and reference it from the load balancer listener so the secret rotation schedule renews the certificate.
ACM auto-renews only certificates it issues when attached to an integrated service like an ALB; imported certificates are not auto-renewed. AWS Certificate Manager provides publicly trusted certificates and handles managed renewal automatically for certificates it issues, provided the certificate is in use on an integrated service such as an Application Load Balancer listener, which removes all manual certificate handling.

Why A is correct: ACM issues publicly trusted certificates and renews them automatically when they are attached to an integrated service such as an ALB listener, meeting the requirement with no manual handling.

Why B is wrong: Self-installing third-party certificates on instances pushes renewal back onto the team and a cron job is exactly the manual handling the requirement rules out.

Why C is wrong: ACM does not renew imported certificates automatically; managed renewal applies only to certificates ACM itself issues, so this still needs manual reimport before expiry.

Why D is wrong: A self-signed certificate is not publicly trusted and Secrets Manager rotation does not issue or renew TLS certificates, so browsers would reject the connection.

Free sampleSecurity and Compliancehard

A CloudOps engineer is told that an IAM role unexpectedly deleted an Amazon S3 bucket policy yesterday, and management wants to know exactly which principal made the call, from which source IP, and when. The account is part of an organisation with an organisation trail already capturing management events to a central S3 bucket. Where should the engineer look to answer these questions with the least new setup?

  • ARun an IAM Access Analyzer scan over the bucket to generate findings that list which external principals deleted the bucket policy and the time of each action.
  • BEnable VPC Flow Logs on the subnets in the account and inspect the captured records to find the principal and the source address that issued the delete call.
  • CQuery the existing CloudTrail management event records in the central S3 bucket for the DeleteBucketPolicy event, reading the principal, source IP and event time fields. Correct
  • DUse the IAM policy simulator against the role to replay the delete request and confirm from the simulation output which statement allowed the action to run.
Use CloudTrail management event records to identify the principal, source IP and time of a past API call during access troubleshooting. CloudTrail captures an event record for each management API call that includes the userIdentity of the caller, the sourceIPAddress and the eventTime, which is exactly the data needed to attribute a past action. Access Analyzer and the policy simulator reason about what a policy permits rather than what happened, and Flow Logs carry only network metadata, so none of them can attribute the historical delete.

Why A is wrong: Access Analyzer reports which principals could access a resource based on policy evaluation; it does not record who actually performed a past API call, so it cannot identify the caller of the delete.

Why B is wrong: VPC Flow Logs capture network traffic metadata such as addresses and ports, but they hold no IAM identity or API action data, so they cannot reveal which principal deleted the policy.

Why C is correct: CloudTrail records each management API call with the calling identity, the source IP address and the timestamp, so querying the captured DeleteBucketPolicy event answers all three questions without new setup.

Why D is wrong: The policy simulator evaluates whether a policy would allow an action in theory; it does not look at historical activity, so it cannot tell you who actually made the call or when.

Free sampleSecurity and Compliancemedium

An AWS Trusted Advisor security check reports that several Amazon EC2 security groups allow unrestricted access from 0.0.0.0/0 to administrative ports such as TCP 22 and TCP 3389. An operations engineer must remediate the finding so that only the corporate office address range can reach these ports. Which action correctly resolves what Trusted Advisor flagged?

  • AAdd a network ACL deny rule on the subnet for TCP 22 and TCP 3389 from 0.0.0.0/0 while leaving the security group rules open, since the stateless ACL takes precedence over the security group.
  • BEdit each flagged security group's inbound rules so that TCP 22 and TCP 3389 are allowed only from the corporate office CIDR range instead of from 0.0.0.0/0, then refresh the check. Correct
  • CMove the affected instances into a private subnet with no internet gateway route, leaving the 0.0.0.0/0 security group rule in place because the subnet now blocks all inbound traffic.
  • DAttach an IAM policy to the instance roles that denies the ec2:AuthorizeSecurityGroupIngress action so no one can widen the rule again, without altering the existing open rule.
Remediating a Trusted Advisor open-security-group finding means narrowing the inbound rule's source from 0.0.0.0/0 to the approved range. Trusted Advisor evaluates the security group inbound rules directly, so the finding clears only when the rule that permits administrative ports from 0.0.0.0/0 is rewritten to allow just the approved corporate CIDR; ACLs, subnet moves and IAM denies leave that rule untouched.

Why A is wrong: A subnet ACL deny does not change the security group rule that Trusted Advisor evaluates, so the check still reports the group as unrestricted and the finding remains.

Why B is correct: Trusted Advisor flags the open-to-the-world administrative ports, so narrowing the source to the office CIDR on the inbound rule removes the exposure and clears the finding on refresh.

Why C is wrong: Trusted Advisor inspects the security group configuration itself, so an open 0.0.0.0/0 rule still triggers the finding even after the instances are moved to a private subnet.

Why D is wrong: Denying a future API call does not remove the already-open inbound rule that Trusted Advisor flagged, so the unrestricted access and the finding both persist.

Other domains in this exam

See also the SOA-C03 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Amazon Web Services. Original, blueprint-aligned practice material only.