AWS Certified Security - Specialty (SCS-C03) — Free Practice Questions
10 free sample questions from a bank of 405, with the correct answers and explanations. No signup required — start practising right now.
1A security engineer is troubleshooting an AWS Lambda function that is named MyLambdaFunction. The function is encountering an error when the function attempts to read the objects in an Amazon S3 bucket that is named DOC-EXAMPLE-BUCKET. The S3 bucket has the following bucket policy:
Which change should the security engineer make to the policy to ensure that the Lambda function can read the bucket objects?
Remove the Condition element. Change the Principal element to the following:
Change the Action element to the following:
Change the Resource element to "arn:aws:s3:::DOC-EXAMPLE- BUCKET/*''.
Change the Resource element to "arn:aws:lambda:::function:MyLambdaFunction". Change the Principal element to the following:
Answer: C
The short version
C — Fix the Resource to the object ARN with wildcard. Bucket-level ARN alone denies object reads, so adding slash-star grants MyLambdaFunction access to contents.
Key concepts in this question
Bucket versus object ARNs: bucket ARN covers the container, object ARN with suffix covers keys inside
Read objects need object-level permission: s3 GetObject evaluates against the key ARN, not just the bucket
Principal and Condition: narrow who and when, but cannot substitute for missing object resource
Lambda resource confusion: Lambda ARNs belong in principals or roles, never as S3 resources
Why C is correct
The function fails reading objects because the policy grants only the bucket itself without its contents. Changing Resource to the DOC-EXAMPLE-BUCKET wildcard path authorizes GetObject on keys the function requests. Other edits leave the resource mismatch that causes the denial.
Why the others are wrong
A. drops conditions and mangles principal without adding object-level resource
B. tweaks actions while the resource still omits object keys
D. places a Lambda ARN as S3 resource, which never matches object requests
SCS-C03 exam tip
Lambda cannot read S3 objects equals check Resource: bucket needs slash-star for keys, bucket alone is not enough.
2The Security team believes that a former employee may have gained unauthorized access to AWS resources sometime in the past 3 months by using an identified access key.
What approach would enable the Security team to find out what the former employee may have done within AWS?
Use the AWS CloudTrail console to search for user activity.
Use the Amazon CloudWatch Logs console to filter CloudTrail data by user.
Use AWS Config to see what actions were taken by the user.
Use Amazon Athena to query CloudTrail logs stored in Amazon S3.
Answer: A
The short version
A — CloudTrail console search reveals the former key activity. Event history retains 90 days of management API events searchable without trail setup. Filtering by access key or user over the past three months exposes what the suspect key did. Logs-based filtering needs configured delivery, Config shows state not API calls, and Athena needs bucket plus tables.
Key concepts in this question
History: Built-in 90-day control-plane record ideal for a three-month lookback with no setup.
Key filter: Scopes to the access key ID isolating former employee actions.
Logs contrast: Powerful for longer retention but depends on prior trail delivery.
Athena contrast: Suits archived S3 trails beyond 90 days with setup overhead.
Why A is correct
The window matches history retention exactly, so opening console history, filtering by the key, and reviewing actions answers the question with minimal latency. Alternatives assume delivered logs or state timelines with extra dependencies. The settled method follows the cited knowledge-center guidance for key activity search.
Why the others are wrong
B. CloudWatch presumes trail to Logs delivery already enabled, never established here.
C. Config shows resource changes, not full API call history for key misuse.
D. Athena needs existing bucket plus tables, overkill for a console-searchable window.
SCS-C03 exam tip
Past 90 days by key equals CloudTrail event history search.
3An AWS account administrator created an IAM group and applied the following managed policy to require that each individual user authenticate using multi-factor authentication:
After implementing the policy, the administrator receives reports that users are unable to perform Amazon EC2 commands using the AWS CLI.
What should the administrator do to resolve this problem while still enforcing multi-factor authentication?
Change the value of aws:MultiFactorAuthPresent to true.
Instruct users to run the aws sts get-session-token CLI command and pass the multi-factor authentication --serial-number and -token-code parameters. Use these resulting values to make API/CLI calls.
Implement federated API/CLI access using SAML 2.0, then configure the identity provider to enforce multi-factor authentication.
Create a role and enforce multi-factor authentication in the role trust policy. Instruct users to run the sts assume-role CLI command and pass --serial-number and --token-code parameters. Store the resulting values in environment variables. Add sts:AssumeRole to NotAction in the policy.
Answer: B
The short version
B — Use get-session-token with MFA for CLI. Temporary credentials carry the MFA context that EC2 API calls require under the group policy.
Key concepts in this question
MFA-enforcing policy: checks aws MultiFactorAuthPresent, true only for MFA-authenticated sessions
Long-lived CLI keys: lack MFA context, so direct EC2 calls fail the condition
Get-session-token: exchanges MFA serial plus code for short-term keys stamped as MFA present
Federation or roles: heavier redesigns, unnecessary when session tokens solve CLI use
Why B is correct
Users blocked on EC2 CLI after MFA enforcement need a way to prove MFA per session. Running get-session-token with serial and token code returns temporary credentials that satisfy the policy condition. That restores CLI function while preserving MFA mandates with least change. This follows AWS security guidance to enforce explicit denies, least privilege, and immutable retention for resilient protection.
Why the others are wrong
A. flips the condition to true globally, effectively bypassing MFA proof
C. rebuilds access via SAML federation, overkill for immediate CLI relief
D. adds roles and NotAction edits, more complex than native session tokens
SCS-C03 exam tip
MFA policy blocks CLI equals get-session-token with serial plus code: fetch MFA-stamped temp keys first.
4A company is using AWS Organizations with the default SCP. The company needs to restrict AWS usage for all AWS accounts that are in a specific OU.
Except for some desired global services, the AWS usage must occur only in the eu-west-1 Region for all accounts in the OU. A security engineer must create an SCP that applies the restriction to existing accounts and any new accounts in the OU.
Which SCP will meet these requirements?
Answer: C
The short version
C — Apply an SCP that denies outside eu-west-1 except global services. Targeted OU policy with region condition plus global-service allows meets existing and new accounts.
Key concepts in this question
Service control policies: guardrails on OU that limit API use regardless of identity permissions
Region condition: aws RequestedRegion or source checks confine actions to eu-west-1
Global services exception: IAM, CloudFront, and similar must remain allowed to function
OU inheritance: attaching at the OU covers current plus future member accounts automatically
Why C is correct
Restricting all regional use to eu-west-1 while permitting required global calls needs a deny-unless-region SCP with an allowlist for global services. Placing it on the OU inherits to existing and newly added accounts. That structure is the settled compliant design for region lock. This follows AWS security guidance to enforce explicit denies, least privilege, and immutable retention for resilient protection.
Why the others are wrong
A. lacks the region-plus-global exception logic for OU-wide restriction
B. misses inheritance-safe region enforcement for new accounts
D. fails to preserve necessary global services while denying other regions
SCS-C03 exam tip
Only one region for an OU equals SCP deny outside region with global allowlist, attached at the OU.
5A company will store sensitive documents in three Amazon S3 buckets based on a data classification scheme of Sensitive,Confidential, and Restricted. The security solution must meet all of the following requirements:
✑ Each object must be encrypted using a unique key.
✑ Items that are stored in the Restricted bucket require two-factor authentication for decryption.
✑ AWS KMS must automatically rotate encryption keys annually.
Which of the following meets these requirements?
Create a Customer Master Key (CMK) for each data classification type, and enable the rotation of it annually. For the ג€Restrictedג€ CMK, define the MFA policy within the key policy. Use S3 SSE-KMS to encrypt the objects.
Create a CMK grant for each data classification type with EnableKeyRotation and MultiFactorAuthPresent set to true. S3 can then use the grants to encrypt each object with a unique CMK.
Create a CMK for each data classification type, and within the CMK policy, enable rotation of it annually, and define the MFA policy. S3 can then create DEK grants to uniquely encrypt each object within the S3 bucket.
Create a CMK with unique imported key material for each data classification type, and rotate them annually. For the ג€Restrictedג€ key material, define the MFA policy in the key policy. Use S3 SSE-KMS to encrypt the objects.
Answer: A
The short version
A — Separate CMKs with rotation plus Restricted MFA satisfy all demands. Each classification needs its own managed key for independent rotation and policy, with yearly rotation enabled. Envelope encryption under SSE-KMS yields a unique data key per object, meeting per-object uniqueness without per-file keys. An MFA condition on the Restricted policy gates its decryption on a second factor.
Key concepts in this question
Per-class key: Isolates Sensitive, Confidential, and Restricted for tailored policy and blast radius.
Rotation: Yearly new backing material while preserving decryption of older data keys.
MFA condition: Requires multi-factor present on Restricted decrypt calls only.
Envelope: S3 requests fresh data key per object under the class CMK and stores it encrypted alongside.
Why A is correct
One CMK per class with rotation gives automatic renewal without re-encryption. SSE-KMS makes every PUT use a fresh data key, satisfying uniqueness through envelope semantics. MFA only on Restricted enforces extra auth exactly where needed. No grant or imported-material variant preserves all three properties at once.
Why the others are wrong
B. grants misuse temporary delegation and cannot persist rotation plus MFA decrypt policy.
C. DEK-grant framing misstates KMS flow and jumbles rotation with MFA placement.
D. imported material with manual rotation adds burden without implementing the policy gate.
SCS-C03 exam tip
Unique per object plus annual plus MFA equals per-class CMK plus SSE-KMS plus MFA.
6What is the effect of the following AWS Key Management Service (AWS KMS} key policy that is attached to a customer managed key?
Amazon WorkMail and Amazon Simple Email Service (Amazon SES) have delegated KMS encrypt and decrypt permissions to the ExampleRole principal in the 111122223333 account.
The ExampleRole principal can transparently encrypt and decrypt email exchanges specifically between ExampleRole and AWS.
The customer managed key can be used for encrypting and decrypting only when the principal is ExampleRole and when the request comes from Amazon WorkMail or Amazon Simple Email Service (Amazon SES) in the specified AWS Region.
The key policy allows Amazon WorkMail or Amazon Simple Email Service (Amazon SES) to encrypt or decrypt on behalf of the ExampleRole for any customer managed key in the account.
Answer: C
The short version
C — Key usable only by ExampleRole via WorkMail or SES in region. Via-service plus caller plus region conditions narrowly scope encryption rights.
Key concepts in this question
KMS grants via services: WorkMail and SES call KMS on behalf of a principal using grants
kms ViaService condition: ties use to requests arriving through named AWS services
CallerAccount and region: further limit which account and geography may use the key
Delegation misunderstanding: services do not gain broad rights to all keys for the role
Why C is correct
The policy effect is conditional: principal must be ExampleRole and the call must come through WorkMail or SES in the stated region. That precise combination explains compliant email encryption without granting general use. Broader readings ignore the via-service and region gates. This follows AWS security guidance to enforce explicit denies, least privilege, and immutable retention for resilient protection.
Why the others are wrong
A. claims services received standing encrypt rights, missing caller and via-service limits
B. suggests transparent any-exchange rights without service and region conditions
D. extends permission to any key in the account, far beyond this single key policy
SCS-C03 exam tip
KMS via-service plus role plus region equals narrow: only that role through that service there may use that key.
7A company wants to deny a specific federated user named Bob access to an Amazon S3 bucket named DOC-EXAMPLE-BUCKET. The company wants to meet this requirement by using a bucket policy. The company also needs to ensure that this bucket policy affects Bob's S3 permissions only. Any other permissions that Bob has must remain intact.
Which policy should the company use to meet these requirements?
Answer: B
The short version
B — Deny Bob explicitly in the bucket policy. Explicit deny on his federated principal for that bucket overrides any allows while leaving other rights intact.
Key concepts in this question
Explicit deny precedence: deny always wins over allow for the same action and resource
Principal-scoped deny: targets only Bob federated identity, preserving others and his non-S3 rights
Bucket policy scope: controls only this bucket, so unrelated permissions stay untouched
Other buckets or allows: remain effective because the deny is resource-specific
Why B is correct
Denying a single federated user on one bucket without disturbing his other access calls for a bucket policy with his ARN in Principal plus deny on bucket actions. That surgical deny blocks DOC-EXAMPLE-BUCKET while his remaining S3 and service rights persist. Broader statements would over-deny. This follows AWS security guidance to enforce explicit denies, least privilege, and immutable retention for resilient protection.
Why the others are wrong
A. fails to isolate Bob on this bucket without affecting wider rights
C. misses the explicit-deny-for-one-principal construction needed here
D. over- or under-scopes, risking collateral denial or no effect on Bob
SCS-C03 exam tip
Deny one federated user on one bucket equals bucket-policy explicit deny for his ARN on that bucket only.
8A company has five AWS accounts and wants to use AWS CloudTrail to log API calls. The log files must be stored in an Amazon S3 bucket that resides in a new account specifically built for centralized services with a unique top-level prefix for each trail. The configuration must also enable detection of any modification to the logs.
Which of the following steps will implement these requirements? (Choose three.)
Create a new S3 bucket in a separate AWS account for centralized storage of CloudTrail logs, and enable ג€Log File Validationג€ on all trails.
Use an existing S3 bucket in one of the accounts, apply a bucket policy to the new centralized S3 bucket that permits the CloudTrail service to use the "s3: PutObject" action and the "s3 GetBucketACL" action, and specify the appropriate resource ARNs for the CloudTrail trails.
Apply a bucket policy to the new centralized S3 bucket that permits the CloudTrail service to use the "s3 PutObject" action and the "s3 GelBucketACL" action, and specify the appropriate resource ARNs for the CloudTrail trails.
Use unique log file prefixes for trails in each AWS account.
Configure CloudTrail in the centralized account to log all accounts to the new centralized S3 bucket.
Enable encryption of the log files by using AWS Key Management Service
Answer: A, C, D
The short version
ACD — New central bucket plus validation plus policy plus prefixes meets goals. Central logging needs a dedicated services-account bucket with validation for tamper detection. A bucket policy must grant CloudTrail write and ACL read for each source ARN, otherwise cross-account delivery fails. Unique top-level prefixes per trail separate five accounts and ease queries and lifecycle.
Key concepts in this question
Central bucket: Isolates audit logs from workload accounts for retention and control.
Validation: Signed digests prove no log was modified after delivery.
Policy: Allows service principal PutObject plus GetBucketACL on exact bucket and prefix ARNs.
Prefixes: Per-trail namespaces prevent collisions and simplify per-account queries.
Why ACD is correct
Creating the validated bucket establishes the protected destination before trails point at it. The precise policy authorizes each member trail to deliver, the cross-account permission that otherwise blocks writes. Distinct prefixes satisfy the unique namespace mandate. Together they implement tamper-evident centralized storage as required.
Why the others are wrong
B. existing bucket contradicts the new-account demand and its policy text targets the wrong destination.
E. central-account trail misplaces configuration, since trails belong in member accounts.
F. KMS encryption is useful but unrequired, so it cannot displace mandated steps.
SCS-C03 exam tip
Central trail integrity equals new bucket plus validation plus policy plus prefixes.
9A company needs a solution to protect critical data from being permanently deleted. The data is stored in Amazon S3 buckets.
The company needs to replicate the S3 objects from the company's primary AWS Region to a secondary Region to meet disaster recovery requirements. The company must also ensure that users who have administrator access cannot permanently delete the data in the secondary Region.
Which solution will meet these requirements?
Configure AWS Backup to perform cross-Region S3 backups. Select a backup vault in the secondary Region. Enable AWS Backup Vault Lock in governance mode for the backups in the secondary Region.
Implement S3 Object Lock in compliance mode in the primary Region. Configure S3 replication to replicate the objects to an S3 bucket in the secondary Region.
Configure S3 replication to replicate the objects to an S3 bucket in the secondary Region. Create an S3 bucket policy to deny the s3:ReplicateDelete action on the S3 bucket in the secondary Region.
Configure S3 replication to replicate the objects to an S3 bucket in the secondary Region. Configure S3 object versioning on the S3 bucket in the secondary Region.
Answer: B
The short version
B — Replicate with Object Lock compliance to block admin deletes. Cross-region copy plus retention prevents permanent removal even by admins.
Key concepts in this question
S3 replication: copies objects to a secondary region for disaster recovery
Object Lock compliance: retention lock that even root cannot shorten during retention
Governance versus compliance: governance allows privileged bypass, compliance does not
Versioning or backup locks alone: leave admin delete paths or lack S3-native immutability
Why B is correct
Protecting from permanent deletion by admins plus secondary-region copy needs immutable replicated copies. Object Lock compliance on replicated objects enforces retention no one can override early. Other designs allow privileged deletes or rely on vault modes mismatched to S3 replication.
Why the others are wrong
A. uses Backup Vault Lock governance for S3 backups, bypassable by admins and not S3 replication
C. denies replicate-delete only, leaving direct object deletes in the second region possible
D. enables versioning without lock, so admins can still delete versions permanently
SCS-C03 exam tip
Admin-proof replicated S3 equals Object Lock compliance plus replication: immutable copies in the second region.
10A security engineer is responding to an incident that is affecting an AWS account. The ID of the account is 1234156789012. The attack created workloads that are distributed across multiple AWS Regions.
The security engineer contains the attack. The security engineer removes all compute and storage resources from all affected Regions. However, the attacker also created an AWS KMS key. The key policy on the KMS key explicitly allows IAM principal kms:* permissions.
The key was scheduled to be deleted the previous day. However, the key is still enabled and usable. The key has an ARN of arn:aws;kms:us-east-2:123456789012:key/mrk-0bb0212cd9864fdea0dcamzo26efb5670. The security engineer must delete the key as quickly as possible.
Which solution will meet this requirement?
Log in to the account by using the account root user credentials. Re-issue the deletion request for the KMS key with a waiting period of 7 days.
Identify the other Regions where the KMS key ID is present and schedule the key for deletion in 7 days.
Update the IAM principal lo allow kms:* permissions on the KMS key ARN. Re-issue the deletion request for the KMS key with a waiting period of 7 days.
Disable the KMS key. Re-issue the deletion request for the KMS key in 30 days.
Answer: B
The short version
B — Delete the multi-region replica key in its other regions. MRK copies persist per region, so scheduling deletion everywhere removes usability quickly.
Key concepts in this question
Multi-region keys: MRK primary plus replicas share key ID material across regions
Per-region deletion: scheduling in one region does not delete replicas elsewhere
Waiting period: 7-day minimum applies per replica after scheduling
Root or disable delays: waste time when the issue is surviving replicas, not permissions
Why B is correct
The key stayed enabled because its MRK replicas in other regions remain usable despite prior scheduling in one place. Identifying replica regions and scheduling deletion there with the short 7-day window eliminates the attacker residual fastest. Reissuing in the same region alone leaves other copies active. This follows AWS security guidance to enforce explicit denies, least privilege, and immutable retention for resilient protection.
Why the others are wrong
A. reissues in the same region while replicas elsewhere stay usable
C. edits IAM permissions that do not remove existing replica key material
D. disables then waits 30 days, slower than 7-day replica deletion
SCS-C03 exam tip
MRK still usable after delete request equals find replicas: schedule deletion in every replica region with 7 days.