Google Professional Cloud Security Engineer Practice Test 54
Question 530
Compliance requires customer authority over key disablement for selected BigQuery tables. Which practical correction or implementation decision is MOST directly appropriate?
- Disable TLS because default at-rest encryption covers all network traffic in transit.
- Inspect resource encryption configuration and authorized key-management operations.
- Test client-certificate absence, certificate rotation and authorized service policies.
- Test key disable and recovery scenario on nonproduction encrypted data before scheduling destruction.
Explanation
Correct Answer
B:Inspect resource encryption configuration and authorized key-management operations. The intended behavior can be checked because supported resources use intended key custody and policy records explain default versus customer-managed behavior.
Incorrect Answers
A:Disable TLS because default at-rest encryption covers all network traffic in transit. This would not correct the Default at-rest encryption versus CMEK workflow. Claiming a customer controls key revocation without configuring CMEK misstates the security model.
C:Test client-certificate absence, certificate rotation and authorized service policies. This could help Mutual TLS for service-to-service identity, but it does not resolve the Default at-rest encryption versus CMEK case. Default encryption at rest is provided by Google Cloud while CMEK adds customer key lifecycle authority.
D:Test key disable and recovery scenario on nonproduction encrypted data before scheduling destruction. This could help KMS key version destruction and restore window, but it does not resolve the Default at-rest encryption versus CMEK case. Default encryption at rest is provided by Google Cloud while CMEK adds customer key lifecycle authority.
Question 531
A training job operates under unexpected default identity and accesses unrelated data. Which practical correction or implementation decision is MOST directly appropriate?
- Verify training service account bindings, data source scope and output artifact ACLs.
- Correlate anomaly times, model logs and downstream API audit events with privacy controls.
- Test tool permission matrix, approval workflow and invocation audit trail.
- Give the training VM a public JSON service account key with project Owner permissions.
Explanation
Correct Answer
A:Verify training service account bindings, data source scope and output artifact ACLs. The intended behavior can be checked because audit evidence attributes reads and writes to intended identity without unrelated dataset access.
Incorrect Answers
B:Correlate anomaly times, model logs and downstream API audit events with privacy controls. This could help AI model monitoring and abuse investigations, but it does not resolve the Vertex AI training and pipeline service identities case. Managed ML pipelines still depend on Google Cloud workload IAM and data protection.
C: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 Vertex AI training and pipeline service identities case. Managed ML pipelines still depend on Google Cloud workload IAM and data protection.
D:Give the training VM a public JSON service account key with project Owner permissions. This would not correct the Vertex AI training and pipeline service identities workflow. Default or shared broad access can disclose unrelated data during experimentation.
Question 532
A business requires audit and redaction of generated responses before they reach external users. An engineer must explain the recommendation for AI output data loss prevention. Which underlying principle is the MOST accurate?
- Locked retention policies can enforce immutability even against privileged operators.
- Prompt-injection defenses require trusted instruction boundaries and downstream authorization, not only content filters.
- AI outputs can leak memorized or retrieved data unless both inputs and outputs are governed.
- Model inference endpoints are APIs that still require authorization, quotas and protected logging.
Explanation
Correct Answer
C:AI outputs can leak memorized or retrieved data unless both inputs and outputs are governed. Therefore the implementation decision is to apply supported privacy filters, data access segmentation and output inspection for sensitive responses.
Incorrect Answers
A:Hiding PII only in the browser DOM does not prevent it from being returned by the API. Locked retention policies can enforce immutability even against privileged operators.
B:Hiding PII only in the browser DOM does not prevent it from being returned by the API. Prompt-injection defenses require trusted instruction boundaries and downstream authorization, not only content filters.
D:Hiding PII only in the browser DOM does not prevent it from being returned by the API. This describes the separate Managed AI endpoint access controls mechanism; the present case instead requires AI output data loss prevention.
Question 533
A response team needs to contain an active compromised workload while preserving forensic evidence. Which practical correction or implementation decision is MOST directly appropriate?
- Exercise credential revocation, incident timeline and recovery in a safe tabletop.
- Validate detection sources, alert routing and incident enrichment.
- Ignore the detection because no application server crashed.
- Test supported detectors and route critical findings to an incident queue.
Explanation
Correct Answer
A:Exercise credential revocation, incident timeline and recovery in a safe tabletop. The intended behavior can be checked because unauthorized access stops and the investigation retains time-stamped evidence and affected-resource inventory.
Incorrect Answers
B:Validate detection sources, alert routing and incident enrichment. This could help Event Threat Detection behavior monitoring, but it does not resolve the Incident response alert routing and containment case. Containment, evidence preservation and recovery are coordinated security operations rather than one dashboard click.
C:Ignore the detection because no application server crashed. This would not correct the Incident response alert routing and containment workflow. Deleting resources and logs before evidence collection can undermine root-cause analysis.
D:Test supported detectors and route critical findings to an incident queue. This could help Security Command Center centralized findings, but it does not resolve the Incident response alert routing and containment case. Containment, evidence preservation and recovery are coordinated security operations rather than one dashboard click.
Question 534
An auditor needs to identify which human impersonated a privileged service account during maintenance. An engineer must explain the recommendation for Service account impersonation and token creation. Which underlying principle is the MOST accurate?
- The permission to act as a service account is distinct from permission to edit the workload configuration.
- Disabling an identity affects all callers using it, not just one label or attached role.
- Impersonation is an explicit IAM privilege and issues temporary credentials for another principal.
- Domain-restricted sharing guardrails constrain identity grants separately from resource firewalls.
Explanation
Correct Answer
C:Impersonation is an explicit IAM privilege and issues temporary credentials for another principal. Therefore the implementation decision is to grant service account token creation only to eligible principals and use impersonation rather than distributing keys.
Incorrect Answers
A:Overbroad Token Creator grants allow unintended principals to act as the target service account. The permission to act as a service account is distinct from permission to edit the workload configuration.
B:Overbroad Token Creator grants allow unintended principals to act as the target service account. Disabling an identity affects all callers using it, not just one label or attached role.
D:Overbroad Token Creator grants allow unintended principals to act as the target service account. Domain-restricted sharing guardrails constrain identity grants separately from resource firewalls.
Question 535
A stolen workload token has IAM read permission and security wants an additional data-exfiltration boundary. Select THREE statements that correctly combine the VPC Service Controls service perimeter method with evidence or its governing principle.
- Add a VPC firewall rule and assume it automatically establishes a VPC Service Controls perimeter.
- Cross-perimeter attempts to restricted services are denied while required in-perimeter workflows work.
- Put the proposed VPC Service Controls configuration in dry-run mode and inspect violations.
- Apply routed network segmentation and firewall policies to restrict hybrid source ranges and workload targets.
- Apply VPC Service Controls perimeters to supported data services with least-privilege ingress and egress policies.
- VPC Service Controls adds a service-oriented exfiltration boundary separate from IAM authorization.
Explanation
Correct Answers
B:Cross-perimeter attempts to restricted services are denied while required in-perimeter workflows work. This is directly observable validation of the VPC Service Controls service perimeter result.
E:Apply VPC Service Controls perimeters to supported data services with least-privilege ingress and egress policies. This is the intended action for VPC Service Controls service perimeter.
F:VPC Service Controls adds a service-oriented exfiltration boundary separate from IAM authorization. This explains why the implementation and evidence fit together.
Incorrect Answers
A:Add a VPC firewall rule and assume it automatically establishes a VPC Service Controls perimeter. This is not part of a safe or effective VPC Service Controls service perimeter design. An IAM role alone does not confine data copied through permitted APIs to untrusted projects.
C:Put the proposed VPC Service Controls configuration in dry-run mode and inspect violations. This serves the different VPC Service Controls dry-run evaluation objective, not the selected VPC Service Controls service perimeter plan. VPC Service Controls adds a service-oriented exfiltration boundary separate from IAM authorization.
D:Apply routed network segmentation and firewall policies to restrict hybrid source ranges and workload targets. This serves the different Hybrid network segmentation policy placement objective, not the selected VPC Service Controls service perimeter plan. VPC Service Controls adds a service-oriented exfiltration boundary separate from IAM authorization.
Question 536
Application pods should reach a database service but not communicate with unrelated workloads. Which approach BEST meets the requirement while respecting the system constraints?
- Enable supported GKE network policy enforcement and define explicit ingress and egress policies.
- Assume each Kubernetes namespace is a complete network firewall without any network policy.
- Use supported VM Manager patch jobs and compliance views with approved maintenance scheduling.
- Scan infrastructure-as-code plans against approved security guardrails and require review before deployment.
Explanation
Correct Answer
A:Enable supported GKE network policy enforcement and define explicit ingress and egress policies. Kubernetes network policies can restrict pod traffic when enforced by the cluster's networking implementation. Verification should demonstrate that only permitted pod-to-pod paths work and unauthorized paths fail during connection tests.
Incorrect Answers
B:Assume each Kubernetes namespace is a complete network firewall without any network policy. This is an unsafe or ineffective shortcut for GKE network policy for pod segmentation. Writing a policy YAML in an unsupported or unenforced cluster does not create isolation.
C:Use supported VM Manager patch jobs and compliance views with approved maintenance scheduling. This primarily addresses VM Manager patch orchestration, not the specific GKE network policy for pod segmentation decision. Here, kubernetes network policies can restrict pod traffic when enforced by the cluster's networking implementation.
D:Scan infrastructure-as-code plans against approved security guardrails and require review before deployment. This primarily addresses Infrastructure as code security validation, not the specific GKE network policy for pod segmentation decision. Here, kubernetes network policies can restrict pod traffic when enforced by the cluster's networking implementation.
Question 537
An operations team must revoke one leaked secret version without revealing other production secrets. Select TWO statements that correctly combine the Secret Manager versioning and least privilege method with evidence or its governing principle.
- Store approved secrets in Secret Manager versions and grant secret access only to intended workloads.
- Apply supported format-preserving encryption with compatible key and tweak handling.
- Configure a supported BigQuery row access policy tied to approved identities and predicates.
- Store production API credentials in a public Dockerfile ARG and assume Secret Manager manages them.
- Authorized workloads retrieve the needed version and revoked versions cannot be accessed.
Explanation
Correct Answers
A:Store approved secrets in Secret Manager versions and grant secret access only to intended workloads. This is the intended action for Secret Manager versioning and least privilege.
E:Authorized workloads retrieve the needed version and revoked versions cannot be accessed. This is directly observable validation of the Secret Manager versioning and least privilege result.
Incorrect Answers
B:Apply supported format-preserving encryption with compatible key and tweak handling. This serves the different Format-preserving encryption data handling objective, not the selected Secret Manager versioning and least privilege plan. Secrets are governed objects with IAM and version lifecycle; runtime variables alone are not a secret vault.
C:Configure a supported BigQuery row access policy tied to approved identities and predicates. This serves the different BigQuery row-level access policy objective, not the selected Secret Manager versioning and least privilege plan. Secrets are governed objects with IAM and version lifecycle; runtime variables alone are not a secret vault.
D:Store production API credentials in a public Dockerfile ARG and assume Secret Manager manages them. This is not part of a safe or effective Secret Manager versioning and least privilege design. Shared plaintext tokens in source repositories remain exposed after workload deployment.
Question 538
A new Compute Engine workload runs under a default service account that might have overly broad legacy permissions. If the relevant Default service account overprivilege safeguard is omitted, which outcome is the MOST directly associated failure?
- Untested constraints may unexpectedly block service integrations and maintenance.
- Excessive default service account grants expand impact if one workload is compromised.
- Broad exemptions undermine a policy intended to eliminate long-lived machine credentials.
- A provider trusting every repository or AWS role can open artifact publishing to attackers.
Explanation
Correct Answer
B:Excessive default service account grants expand impact if one workload is compromised. Default service account identity and its IAM grants are separate from the VM operating system login. The corrective approach is to inspect service account attachments and IAM permissions before and after workload migration.
Incorrect Answers
A:Untested constraints may unexpectedly block service integrations and maintenance. This failure pattern belongs chiefly to Custom organization policy on resource attributes; it is not the failure most directly tested in the Default service account overprivilege situation. Excessive default service account grants expand impact if one workload is compromised.
C:Broad exemptions undermine a policy intended to eliminate long-lived machine credentials. This failure pattern belongs chiefly to Service account key creation organization policy; it is not the failure most directly tested in the Default service account overprivilege situation. Excessive default service account grants expand impact if one workload is compromised.
D:This failure pattern belongs chiefly to Workload Identity Federation for CI systems; it is not the failure most directly tested in the Default service account overprivilege situation.
Question 539
A new engineer must inherit one reviewed viewer role through the operations group. After the recommended Group-based IAM authorization change is implemented, which observation MOST directly verifies that it works?
- The intended task succeeds without granting administrative privileges to unrelated projects.
- Onboarding and removal change effective access in the intended projects without unrelated grants.
- The identified rule explains successful and failed operations for the exact principal and resource.
- Old privileged sessions expire according to policy and authorized workers can reauthenticate.
Explanation
Correct Answer
B:Onboarding and removal change effective access in the intended projects without unrelated grants. This directly tests whether the Group-based IAM authorization goal was met. Group-based IAM reduces binding sprawl but relies on trustworthy group membership governance.
Incorrect Answers
A:The intended task succeeds without granting administrative privileges to unrelated projects. This is evidence about IAM least-privilege role scoping, not direct proof of the required Group-based IAM authorization behavior. The relevant check is that onboarding and removal change effective access in the intended projects without unrelated grants.
C:The identified rule explains successful and failed operations for the exact principal and resource. This is evidence about IAM effective permission troubleshooting, not direct proof of the required Group-based IAM authorization behavior. The relevant check is that onboarding and removal change effective access in the intended projects without unrelated grants.
D:Old privileged sessions expire according to policy and authorized workers can reauthenticate. This is evidence about Session duration and reauthentication policy, not direct proof of the required Group-based IAM authorization behavior. The relevant check is that onboarding and removal change effective access in the intended projects without unrelated grants.
Question 540
A reviewer needs evidence before reducing excessive project permissions. An engineer must explain the recommendation for Policy Intelligence permission recommendations. Which underlying principle is the MOST accurate?
- Separation of duties limits privilege escalation and tampering with oversight controls.
- Directory synchronization maps authoritative identity sources into managed cloud accounts rather than granting IAM roles itself.
- Just-in-time privileged access reduces standing administrative exposure and preserves an approval record.
- Usage-based IAM recommendations can highlight excess privilege but need contextual validation.
Explanation
Correct Answer
D:Usage-based IAM recommendations can highlight excess privilege but need contextual validation. Therefore the implementation decision is to evaluate IAM Recommender and Policy Intelligence findings against actual workload use and required operations.
Incorrect Answers
A:Separation of duties limits privilege escalation and tampering with oversight controls. This describes the separate IAM separation of security duties mechanism; the present case instead requires Policy Intelligence permission recommendations. Blind recommendation acceptance may remove low-frequency disaster recovery permissions.
B:Directory synchronization maps authoritative identity sources into managed cloud accounts rather than granting IAM roles itself. This describes the separate Cloud Identity directory synchronization mechanism; the present case instead requires Policy Intelligence permission recommendations. Blind recommendation acceptance may remove low-frequency disaster recovery permissions.
C:Just-in-time privileged access reduces standing administrative exposure and preserves an approval record. This describes the separate Privileged Access Manager just-in-time approval mechanism; the present case instead requires Policy Intelligence permission recommendations. Blind recommendation acceptance may remove low-frequency disaster recovery permissions.