Google Professional Cloud Security Engineer Practice Test 37
Question 360
An audit reviewer needs assurance that key policy authors cannot automatically read ciphertext content. Which practical correction or implementation decision is MOST directly appropriate?
- Test supported machine family, image compatibility and confidential mode evidence.
- Review IAM roles and exercise encrypt or decrypt denial with representative identities.
- Inventory encrypted resource dependencies and test decrypt before retiring old versions.
- Give every KMS administrator an organization Owner role for convenience.
Explanation
Correct Answer
B:Review IAM roles and exercise encrypt or decrypt denial with representative identities. The intended behavior can be checked because test users can administer key settings or decrypt only according to their distinct approved roles.
Incorrect Answers
A:Test supported machine family, image compatibility and confidential mode evidence. This could help Confidential VM memory protection, but it does not resolve the Cloud KMS crypto key role separation case. Cloud KMS separates administrative operations from cryptographic actions through IAM permissions.
C:Inventory encrypted resource dependencies and test decrypt before retiring old versions. This could help Customer-managed key rotation policy, but it does not resolve the Cloud KMS crypto key role separation case. Cloud KMS separates administrative operations from cryptographic actions through IAM permissions.
D:Give every KMS administrator an organization Owner role for convenience. This would not correct the Cloud KMS crypto key role separation workflow. Combining key admin and decrypter grants broadly defeats separation of duties.
Question 361
A departing user appears in a historical group report even though their active access should be removed. Select TWO statements that correctly combine the Group-based provisioning audit trail method with evidence or its governing principle.
- Configure Cloud Identity third-party SAML SSO with validated signing certificate and identity attributes.
- The group audit trail explains each addition or removal and no unapproved current member remains.
- Compare authoritative HR events, group membership changes and effective IAM grants using audit evidence.
- Configure Workforce Identity Federation pools, identity provider mappings and least-privilege IAM bindings.
- Assume a group name containing restricted guarantees its members cannot access broader projects.
Explanation
Correct Answers
B:The group audit trail explains each addition or removal and no unapproved current member remains. This is directly observable validation of the Group-based provisioning audit trail result.
C:Compare authoritative HR events, group membership changes and effective IAM grants using audit evidence. This is the intended action for Group-based provisioning audit trail.
Incorrect Answers
A:Configure Cloud Identity third-party SAML SSO with validated signing certificate and identity attributes. This serves the different SAML enterprise single sign-on objective, not the selected Group-based provisioning audit trail plan. Identity groups and IAM inheritance require separate checks for effective access.
D:Configure Workforce Identity Federation pools, identity provider mappings and least-privilege IAM bindings. This serves the different Workforce Identity Federation for external workers objective, not the selected Group-based provisioning audit trail plan. Identity groups and IAM inheritance require separate checks for effective access.
E:Assume a group name containing restricted guarantees its members cannot access broader projects. This is not part of a safe or effective Group-based provisioning audit trail design. Checking only successful SSO sessions overlooks excess permissions inherited through groups.
Question 362
A partner's branch connects over VPN but should reach only one application subnet. After the recommended Hybrid network segmentation policy placement change is implemented, which observation MOST directly verifies that it works?
- Packet tests and firewall logs show expected policy evaluation for allowed and blocked flows.
- Only approved cross-boundary calls succeed and unrelated principals remain denied.
- The connected branch accesses only the approved service and unrelated project traffic is denied.
- Tests show the caller has both sufficient resource permissions and a permitted perimeter path.
Explanation
Correct Answer
C:The connected branch accesses only the approved service and unrelated project traffic is denied. This directly tests whether the Hybrid network segmentation policy placement goal was met. Hybrid connectivity creates a route path but application and subnet firewall policy still governs permitted flows.
Incorrect Answers
A:Packet tests and firewall logs show expected policy evaluation for allowed and blocked flows. This is evidence about Hierarchical Cloud NGFW policy ordering, not direct proof of the required Hybrid network segmentation policy placement behavior. The relevant check is that the connected branch accesses only the approved service and unrelated project traffic is denied.
B:Only approved cross-boundary calls succeed and unrelated principals remain denied. This is evidence about VPC Service Controls ingress egress policy, not direct proof of the required Hybrid network segmentation policy placement behavior. The relevant check is that the connected branch accesses only the approved service and unrelated project traffic is denied.
D:Tests show the caller has both sufficient resource permissions and a permitted perimeter path. This is evidence about VPC Service Controls versus IAM authorization, not direct proof of the required Hybrid network segmentation policy placement behavior. The relevant check is that the connected branch accesses only the approved service and unrelated project traffic is denied.
Question 363
Security wants new virtual machines to avoid unnecessary external IP addresses. Which approach BEST meets the requirement while respecting the system constraints?
- Define and enforce an applicable custom organization policy constraint with supported resource rules.
- Use supported organization policy domain-restricted sharing controls with approved identities or exceptions.
- Remove external IPs from a spreadsheet and assume new VMs can no longer request them.
- Configure the supported Compute Engine external-IP restriction organization policy and review project exceptions.
Explanation
Correct Answer
D:Configure the supported Compute Engine external-IP restriction organization policy and review project exceptions. Organization policy constraints enforce supported resource configuration beyond individual IAM grants. Verification should demonstrate that unapproved VMs cannot acquire public IP addresses while approved exceptions are recorded.
Incorrect Answers
A:Define and enforce an applicable custom organization policy constraint with supported resource rules. This primarily addresses Custom organization policy on resource attributes, not the specific Restricting external IPv4 addresses through org policy decision. Here, organization policy constraints enforce supported resource configuration beyond individual IAM grants.
B:Use supported organization policy domain-restricted sharing controls with approved identities or exceptions. This primarily addresses Restricting allowed IAM member domains, not the specific Restricting external IPv4 addresses through org policy decision. Here, organization policy constraints enforce supported resource configuration beyond individual IAM grants.
C:Remove external IPs from a spreadsheet and assume new VMs can no longer request them. This is an unsafe or ineffective shortcut for Restricting external IPv4 addresses through org policy. A policy with broad project exemption may restore publicly reachable attack paths.
Question 364
A developer submits a pull request adding a dependency with a publicly disclosed critical flaw. If the relevant Continuous integration dependency security gate safeguard is omitted, which outcome is the MOST directly associated failure?
- Ignoring scan results until after deployment reduces time to remediate vulnerable releases.
- Updating a policy does not automatically retroactively remove every existing running revision.
- Signing only a repository name cannot prove integrity of a changed image digest.
- Giving all developers logging administrator and private-log access may expose sensitive audit metadata.
Explanation
Correct Answer
A:Ignoring scan results until after deployment reduces time to remediate vulnerable releases. Shift-left scanning detects known issues during delivery but needs a managed exception process. The corrective approach is to test a known vulnerable fixture and verify developer feedback and gate behavior.
Incorrect Answers
B:This failure pattern belongs chiefly to Binary Authorization Cloud Run revision policy; it is not the failure most directly tested in the Continuous integration dependency security gate situation.
C:This failure pattern belongs chiefly to Binary Authorization attestor chain of custody; it is not the failure most directly tested in the Continuous integration dependency security gate situation.
D:This failure pattern belongs chiefly to Least privilege access to audit logs; it is not the failure most directly tested in the Continuous integration dependency security gate situation.
Question 365
A vendor must access APIs only from a defined network context rather than all locations. Testing exposes this failure: Defining an unused access level alone does not restrict Google Cloud API access. Which remediation should be applied FIRST?
- Bind the access level to the intended access policy and test allowed and denied contexts.
- Create a named access level then assume every project automatically enforces it.
- Bind a suitable predefined or reviewed custom IAM role at the narrowest supported scope.
- Enable uniform bucket-level access where appropriate and manage object access through IAM.
Explanation
Correct Answer
A:Bind the access level to the intended access policy and test allowed and denied contexts. Confirmation requires seeing that requests that satisfy the access level are eligible while unmatched contexts are rejected by the consuming policy.
Incorrect Answers
B:Create a named access level then assume every project automatically enforces it. This retains the Access Context Manager access levels problem. Access levels describe contextual trust conditions and must be attached to an enforcing service policy.
C:Bind a suitable predefined or reviewed custom IAM role at the narrowest supported scope. It addresses IAM least-privilege role scoping, not the diagnosed Access Context Manager access levels failure. Defining an unused access level alone does not restrict Google Cloud API access.
D:Enable uniform bucket-level access where appropriate and manage object access through IAM. It addresses Cloud Storage ACL versus IAM with uniform access, not the diagnosed Access Context Manager access levels failure. Defining an unused access level alone does not restrict Google Cloud API access.
Question 366
A business requires audit and redaction of generated responses before they reach external users. Which practical correction or implementation decision is MOST directly appropriate?
- Perform adversarial data-exposure tests and verify effective data classification policies.
- Test tool permission matrix, approval workflow and invocation audit trail.
- Turn off query logging and assume that alone removes private identifiers from all model outputs.
- Test data access, job execution identity and output artifact permissions before training.
Explanation
Correct Answer
A:Perform adversarial data-exposure tests and verify effective data classification policies. The intended behavior can be checked because test prompts do not expose protected identifiers and redaction events remain auditable.
Incorrect Answers
B:Test tool permission matrix, approval workflow and invocation audit trail. This could help Gemini Enterprise Agent Platform tool permission design, but it does not resolve the AI output data loss prevention case. AI outputs can leak memorized or retrieved data unless both inputs and outputs are governed.
C:Turn off query logging and assume that alone removes private identifiers from all model outputs. This would not correct the AI output data loss prevention workflow. Hiding PII only in the browser DOM does not prevent it from being returned by the API.
D:Test data access, job execution identity and output artifact permissions before training. This could help AI training data isolation and least privilege, but it does not resolve the AI output data loss prevention case. AI outputs can leak memorized or retrieved data unless both inputs and outputs are governed.
Question 367
A training job operates under unexpected default identity and accesses unrelated data. An engineer must explain the recommendation for Vertex AI training and pipeline service identities. Which underlying principle is the MOST accurate?
- Prompt-injection defenses require trusted instruction boundaries and downstream authorization, not only content filters.
- Managed ML pipelines still depend on Google Cloud workload IAM and data protection.
- Imported key material requires compatible algorithms and a protected import workflow.
- Model inference endpoints are APIs that still require authorization, quotas and protected logging.
Explanation
Correct Answer
B:Managed ML pipelines still depend on Google Cloud workload IAM and data protection. Therefore the implementation decision is to attach a dedicated supported service account and restrict managed training pipeline permissions.
Incorrect Answers
A:Default or shared broad access can disclose unrelated data during experimentation. Prompt-injection defenses require trusted instruction boundaries and downstream authorization, not only content filters.
C:Default or shared broad access can disclose unrelated data during experimentation. Imported key material requires compatible algorithms and a protected import workflow.
D:Default or shared broad access can disclose unrelated data during experimentation. This describes the separate Managed AI endpoint access controls mechanism; the present case instead requires Vertex AI training and pipeline service identities.
Question 368
A regulated program needs to verify which Google Cloud services are in scope of a specific third-party assurance report. An engineer must explain the recommendation for Vendor shared-control contract and service eligibility. Which underlying principle is the MOST accurate?
- Rotating a key typically changes the primary version for new operations and does not necessarily re-encrypt existing data.
- Provider attestations cover specified services and periods, not every customer configuration or integration.
- Healthcare privacy controls require data minimization, authorization and auditability regardless of infrastructure encryption.
- Audit frameworks assess the operation of controls, not only the existence of a vendor badge.
Explanation
Correct Answer
B:Provider attestations cover specified services and periods, not every customer configuration or integration. Therefore the implementation decision is to compare official service compliance scope, customer contract terms and workload-specific requirements.
Incorrect Answers
A:Assuming a vendor certificate transfers regulatory responsibility to the customer workload may produce audit gaps. Rotating a key typically changes the primary version for new operations and does not necessarily re-encrypt existing data.
C:Assuming a vendor certificate transfers regulatory responsibility to the customer workload may produce audit gaps. This describes the separate Healthcare data privacy segmentation mechanism; the present case instead requires Vendor shared-control contract and service eligibility.
D:Assuming a vendor certificate transfers regulatory responsibility to the customer workload may produce audit gaps. Audit frameworks assess the operation of controls, not only the existence of a vendor badge.
Question 369
An administrator needs to know when a Binary Authorization policy is evaluated for Cloud Run jobs versus services. Testing exposes this failure: Updating a policy does not automatically retroactively remove every existing running revision. Which remediation should be applied FIRST?
- Test revision deployment, job execution and documented emergency break-glass logging.
- Enable supported container vulnerability scanning and connect findings to release review.
- Assume changing a Binary Authorization policy instantly rewrites or terminates all existing Cloud Run revisions.
- Bind workload-specific Kubernetes subjects to approved Google Cloud resource roles.
Explanation
Correct Answer
A:Test revision deployment, job execution and documented emergency break-glass logging. Confirmation requires seeing that noncompliant new service revisions or job executions fail at their respective enforcement points.
Incorrect Answers
B:Enable supported container vulnerability scanning and connect findings to release review. It addresses Artifact Registry vulnerability scanning policy, not the diagnosed Binary Authorization Cloud Run revision policy failure. Updating a policy does not automatically retroactively remove every existing running revision.
C:Assume changing a Binary Authorization policy instantly rewrites or terminates all existing Cloud Run revisions. This retains the Binary Authorization Cloud Run revision policy problem. Cloud Run revision admission and job execution have distinct Binary Authorization evaluation timing.
D:Bind workload-specific Kubernetes subjects to approved Google Cloud resource roles. It addresses GKE Workload Identity for deployment runtime, not the diagnosed Binary Authorization Cloud Run revision policy failure. Updating a policy does not automatically retroactively remove every existing running revision.
Question 370
A deployed GKE pod needs to access one Cloud Storage bucket but not node-level credentials. After the recommended GKE Workload Identity for deployment runtime change is implemented, which observation MOST directly verifies that it works?
- The selected instances install approved updates and show patch compliance and failures.
- Only the intended pods access the bucket without exporting long-lived private service keys.
- An intentionally unsafe plan fails the gate while a least-privilege revision passes.
- Expected source-destination flow metadata appears without claiming every packet payload is captured.
Explanation
Correct Answer
B:Only the intended pods access the bucket without exporting long-lived private service keys. This directly tests whether the GKE Workload Identity for deployment runtime goal was met. Runtime workload identity scopes cloud permissions to pods and service accounts rather than entire hosts.
Incorrect Answers
A:The selected instances install approved updates and show patch compliance and failures. This is evidence about VM Manager patch orchestration, not direct proof of the required GKE Workload Identity for deployment runtime behavior. The relevant check is that only the intended pods access the bucket without exporting long-lived private service keys.
C:An intentionally unsafe plan fails the gate while a least-privilege revision passes. This is evidence about Infrastructure as code security validation, not direct proof of the required GKE Workload Identity for deployment runtime behavior. The relevant check is that only the intended pods access the bucket without exporting long-lived private service keys.
D:Expected source-destination flow metadata appears without claiming every packet payload is captured. This is evidence about VPC Flow Logs traffic analysis, not direct proof of the required GKE Workload Identity for deployment runtime behavior. The relevant check is that only the intended pods access the bucket without exporting long-lived private service keys.