# Free SCS-C03 practice questions

> 10 free SCS-C03 practice questions with the correct answer and an explanation.

Page: https://devopscodelab.com/exams/aws-scs-c03/free-practice-questions/
Practice exams: https://devopscodelab.com/exams/aws-scs-c03/

### Question 1 of 10

A financial company uses AWS Organizations with 120 accounts across four Regions. The security team wants a single place to view Security Hub findings from all accounts and Regions, with security standards and controls enabled consistently in every new account without manual steps per account or Region. Which combination of configuration achieves this?

- A. Enable Security Hub in every account and Region with CloudFormation StackSets that also enable the standards, then send member invitations from a central administrator account
- B. Designate a delegated administrator, turn on auto-enable for new accounts in the home Region, and rely on finding replication to cover the other three Regions
- C. Designate a delegated administrator, enable cross-Region aggregation to a home Region, and use central configuration policies associated with the organization root
- D. Have each account forward Security Hub findings through Amazon EventBridge to a central Amazon Security Lake and query the combined findings with Athena from the security account

**Answer:** C

**Explanation:** Central configuration lets the delegated administrator define configuration policies (which standards and controls are enabled) and apply them to the root, OUs, or accounts across all linked Regions, while cross-Region aggregation brings findings into one home Region. Forwarding to Security Lake gives queryable storage but neither a unified Security Hub view nor consistent standards. StackSets plus invitations is the legacy, labor-intensive model and does not cover organization members automatically. Auto-enable in one Region does not configure the other Regions, and aggregation does not enable anything.

### Question 2 of 10

A company runs containerized workloads on Amazon ECS with AWS Fargate across 30 accounts in AWS Organizations. The security team wants to detect suspicious process executions, reverse shells, and crypto-mining binaries running inside the tasks, with findings centralized in the security tooling account. Which solution meets these requirements with the LEAST operational overhead?

- A. Make the security tooling account the GuardDuty delegated administrator and enable Runtime Monitoring for ECS Fargate with automated agent management for all members
- B. Delegate Amazon Inspector to the security tooling account and enable enhanced scanning with continuous rescans for every Amazon ECR repository used by the Fargate tasks
- C. Enable VPC Flow Logs on every task subnet in all accounts and use Amazon Detective in the security tooling account to investigate anomalous outbound connections from the tasks
- D. Add a third-party runtime security sidecar to every task definition through a shared base template, and forward its process and syscall logs to a central Amazon OpenSearch Service domain for detection rules

**Answer:** A

**Explanation:** GuardDuty Runtime Monitoring observes operating-system-level activity (process, file, and network events) inside Fargate tasks, and with automated agent management the delegated administrator rolls it out organization-wide without touching task definitions. Flow logs and Detective show network metadata only and cannot see process executions. A sidecar in every task definition plus self-built detection rules in OpenSearch is far more operational work. Inspector continuous scanning finds known vulnerabilities in images, not malicious behavior while the tasks run.

### Question 3 of 10

A compromised EC2 instance has an established outbound TCP session to an attacker's server. The responder replaced the instance's security group with one that has no inbound or outbound rules, but the session is still active. Why, and what should the responder do to cut the session immediately?

- A. A security group without rules still allows all outbound traffic by default, so the responder should add an explicit outbound deny rule
- B. Security group changes can take up to an hour to propagate to every host, so the responder should wait and confirm with VPC Flow Logs
- C. The network interface caches security group rules, so the responder should detach and reattach the isolation group to force a refresh
- D. Security groups are stateful and do not interrupt tracked connections; add network ACL deny rules for the instance's traffic to drop it

**Answer:** D

**Explanation:** Existing tracked flows continue after a security group change because security groups are stateful; network ACLs are stateless and evaluated per packet, so a deny rule stops the session immediately (ideally with the instance isolated in its own subnet, or with the NACL change scoped carefully). Security group changes apply within seconds, not an hour. A security group with no outbound rules allows no outbound traffic, and security groups do not support deny rules anyway. Reattaching a group does not reset tracked connections.

### Question 4 of 10

A company wants all newly created EBS snapshots and AMIs in every account to be impossible to share publicly, regardless of user permissions, while allowing sharing with specific accounts. Which approach provides this with the LEAST ongoing effort?

- A. Remove ec2:ModifySnapshotAttribute and ec2:ModifyImageAttribute from every IAM role in every account
- B. Create an AWS Config rule for public snapshots and AMIs with automatic remediation that removes the public permission
- C. Encrypt every snapshot and AMI with the AWS managed key aws/ebs, because those cannot be shared publicly
- D. Enable EBS snapshot and AMI block public access through an Organizations declarative policy

**Answer:** D

**Explanation:** Block public access settings for snapshots and AMIs prevent public sharing while still allowing account-specific sharing, and a declarative policy enforces them across the organization including new accounts. Config remediation is reactive and leaves a window of exposure. Snapshots encrypted with the AWS managed key cannot be shared with any other account, which breaks the requirement. Removing the permissions also blocks legitimate account-specific sharing.

### Question 5 of 10

A company wants administrators to connect to Linux EC2 instances in private subnets without bastion hosts, without opening inbound ports, and with every session's commands logged to S3. Which solution meets all requirements?

- A. AWS Client VPN with certificate authentication into the VPC, and SSH to the instances with keys stored in Secrets Manager
- B. A hardened bastion host in a public subnet with SSH restricted to the corporate IP range and the CloudWatch agent shipping shell history
- C. EC2 Instance Connect Endpoint in the private subnet, with short-lived SSH keys pushed per session by the console
- D. Systems Manager Session Manager with session logging to S3 and VPC interface endpoints for Systems Manager

**Answer:** D

**Explanation:** Session Manager uses the SSM Agent's outbound connection, so no inbound ports are needed, and it can log session data to S3; interface endpoints let private subnets reach Systems Manager without internet access. A bastion requires inbound ports and is itself a host to manage. Client VPN plus SSH needs open SSH ports and has no built-in command logging. EC2 Instance Connect Endpoint needs inbound SSH from the endpoint and does not log commands.

### Question 6 of 10

A company wants to prevent principals outside its organization from accessing any S3 bucket, KMS key, or Secrets Manager secret in its accounts, even if a developer writes a resource policy that grants such access. Which policy type is designed for this?

- A. A tag policy that requires a DataPerimeter tag on every bucket, key, and secret and enforces its allowed values
- B. An RCP that denies access unless aws:PrincipalOrgID matches, with exceptions for AWS service principals
- C. A permissions boundary on every IAM role in member accounts that denies access to resources outside the organization
- D. An SCP attached to the root that denies s3:PutBucketPolicy, kms:PutKeyPolicy, and secretsmanager:PutResourcePolicy to developers

**Answer:** B

**Explanation:** Resource control policies apply to resources in member accounts and cap what resource policies can grant, so external principals are blocked regardless of individual policies; conditions such as aws:PrincipalIsAWSService keep service access working. An SCP limits the company's own principals and does not stop external callers, and blocking policy edits breaks legitimate work. Tag policies only standardize tags. Permissions boundaries limit identities in the account, not outside callers.

### Question 7 of 10

A company's GitHub Actions workflows deploy to AWS using long-term access keys stored as repository secrets. Security wants to eliminate long-term keys and ensure that only workflows from the main branch of one specific repository can deploy to production. What should be implemented?

- A. A GitHub OIDC identity provider in IAM and a deploy role whose trust policy checks aud and sub for the repo and main branch
- B. IAM Roles Anywhere, with a client certificate and private key stored as repository secrets for the workflows
- C. An IAM user per repository whose access keys a Lambda function rotates weekly and writes back to GitHub secrets
- D. An Amazon Cognito user pool that issues tokens to GitHub runners, exchanged for credentials through an identity pool

**Answer:** A

**Explanation:** AssumeRoleWithWebIdentity with GitHub's OIDC provider issues short-lived credentials, and conditions on the sub claim restrict access to one repository and branch. Roles Anywhere would require storing a private key as a secret, which is again a long-lived credential. Rotated IAM user keys are still long-term secrets. A Cognito user pool adds a separate identity store and cannot bind access to a repository and branch the way the GitHub token claims do.

### Question 8 of 10

A security engineer wants to delete a customer managed KMS key that appears unused, but must be certain no application still depends on it before the key is gone permanently. What is the safest approach?

- A. Rotate the key on demand and then delete the previous key material once rotation has completed
- B. Delete the key's alias and wait a week for applications to report errors before scheduling deletion of the key
- C. Disable the key or schedule deletion with the longest waiting period, and alarm on CloudTrail attempts to use it
- D. Delete the key immediately, because unused customer managed keys add monthly cost and attack surface

**Answer:** C

**Explanation:** Disabling the key or scheduling deletion with a waiting period allows recovery if something fails, and alarming on attempted use reveals hidden dependencies before deletion completes. Removing an alias does not reveal applications that use the key ID or ARN. KMS keys cannot be deleted immediately, and deleting without checks risks permanent data loss. Older key material cannot be deleted separately.

### Question 9 of 10

A company runs PostgreSQL on Amazon RDS and must ensure that every client connection uses TLS with certificate verification. What should be configured?

- A. Set rds.force_ssl to 1 in the parameter group, and have clients use sslmode=verify-full with the RDS CA bundle
- B. Enable encryption at rest for the instance with a customer managed KMS key and enforce key rotation
- C. Enable IAM database authentication and require all application users to log in with authentication tokens
- D. Place the database in a private subnet with a security group that allows only the application servers

**Answer:** A

**Explanation:** rds.force_ssl rejects non-TLS connections, and verify-full on clients validates the server certificate against the RDS CA, preventing man-in-the-middle attacks. Private subnets restrict network reach but not encryption. Encryption at rest does not affect transport. IAM authentication requires TLS for token logins, but it does not force other connections to use TLS or verify certificates.

### Question 10 of 10

An organization's root has the default FullAWSAccess SCP. An OU named Sandbox has an SCP that allows only ec2:* and s3:*, and FullAWSAccess was detached from that OU. An account under Sandbox has FullAWSAccess attached directly. Which actions can principals in that account perform, assuming broad IAM permissions?

- A. Only EC2 and S3 actions, because SCPs must allow an action at every level from the root to the account
- B. No actions, because the OU's allow list and the account's FullAWSAccess conflict and default to deny
- C. All actions, because FullAWSAccess attached directly to the account takes precedence over the OU's SCP
- D. All actions except EC2 and S3, because the OU's SCP is evaluated as an exclusion list for its accounts

**Answer:** A

**Explanation:** SCP allow statements must permit an action at each level of the hierarchy, so the OU's allow list limits the account even though FullAWSAccess is attached directly. The OU allows EC2 and S3, not everything else. SCPs at different levels intersect rather than conflict into a deny-all. Direct attachment does not bypass parent restrictions.

Ready for a timed run? Take the free 15-question SCS-C03 mini exam: https://devopscodelab.com/app/exams/aws-scs-c03/mini/

DevOpsCode Lab is not affiliated with AWS. SCS-C03 and related names are trademarks of their owners. We write every question ourselves; these are not real exam questions.
