GPSO
Free GPSO practice questions
10 questions from our GPSO practice tests, each with the correct answer and a written explanation. Read them with the answers shown, or switch to quiz mode and answer first.
Question 1 of 10, Platform Operations
A company uses Microsoft Entra ID as its corporate identity provider and does not sync users into Cloud Identity. Analysts must sign in to Google Security Operations with their Entra ID credentials and receive permissions through Google Cloud IAM roles. What should the platform engineer configure?
- AA workload identity pool trusting Entra ID tokens, then grant the SecOps roles to the federated service account used by each analyst
- BA workforce identity pool and OIDC or SAML provider for Entra ID, then grant SecOps IAM roles to the pool's principal sets Correct answer
- CAn Entra ID enterprise application that issues API keys for SecOps, with each analyst storing their key in Secret Manager for sign-in
- DGoogle Cloud Directory Sync to copy Entra ID users into Cloud Identity, then grant the SecOps IAM roles to the synchronized groups
Answer and explanation: Workforce Identity Federation lets users of an external IdP such as Entra ID authenticate to Google Cloud services, including Google SecOps, without being synced into Cloud Identity, and IAM roles are granted to workforce pool principals. Workload identity pools are for machine workloads, not interactive human users. Directory Sync would work technically but violates the stated decision not to sync users. API keys are not a sign-in method for the SecOps console and do not carry user identity.
Question 2 of 10, Platform Operations
A global company wants to deploy Cloud IDS to inspect traffic for production workloads running in three regions within a single Shared VPC. Leadership wants coverage of all production subnets without inspecting development traffic, which shares the same VPC. Which design is MOST appropriate?
- ACreate Cloud IDS endpoints in each production zone and packet mirroring policies that select only production subnets or tagged instances Correct answer
- BDeploy Cloud IDS endpoints only in the development subnets, because production traffic is already protected by Cloud Armor
- CEnable Cloud IDS at the organization level, which inspects every subnet automatically without packet mirroring or endpoints
- DCreate one Cloud IDS endpoint in a single region and mirror traffic from all subnets in every region to it, keeping endpoint count and hourly costs to a minimum
Answer and explanation: Cloud IDS endpoints are zonal resources that receive traffic through packet mirroring policies, which can target specific subnets, instances, or tags, so production can be covered while development traffic is excluded. Mirroring must stay within the endpoint's region, so a single endpoint cannot serve all regions. There is no organization-wide switch that inspects traffic without endpoints and mirroring. Cloud Armor protects external load balancers and does not inspect internal VPC traffic.
Question 3 of 10, Data Management
A Google SecOps rule joins EDR process events with DHCP logs to attribute activity to laptops. Detections show only IP addresses even though asset context from the CMDB contains hostnames and MAC addresses. DHCP logs are parsed correctly. What is the MOST likely reason enrichment is not adding the hostname?
- AAsset aliasing cannot link the IP to the host because identifiers are missing or mapped to unexpected UDM fields in one source Correct answer
- BThe CMDB feed must be converted to event data, because entity context is only used for enrichment when it arrives as UDM events
- CEnrichment works only for user entities, so asset fields such as hostname and MAC address are never added to detections automatically
- DHostname enrichment requires the detection rule to run as a retrohunt, because live rules only evaluate fields present in the raw logs
Answer and explanation: Google SecOps uses aliasing to link IP addresses, MAC addresses, hostnames, and asset IDs over time, and enrichment fails when those identifiers are absent or mapped to unexpected fields in the sources. Entity context feeds are supposed to arrive as entity data, not events. Enrichment applies to assets as well as users. Live rules and retrohunts use the same enriched data, so running a retrohunt would not change the result.
Question 4 of 10, Threat Hunting. Choose 2.
A hunter forms a hypothesis that an attacker is abusing a compromised developer account to enumerate secrets. Which combination of data sources gives the STRONGEST evidence to confirm or refute the hypothesis in Google Cloud? (Choose two.)
- ACloud Billing reports showing the project's Secret Manager spend for the current month
- BVPC Flow Logs showing traffic volume between subnets in the developer's project
- CAuthentication events showing the developer's sign-ins with source IP addresses and user agents Correct answer
- DSecret Manager Data Access audit logs showing AccessSecretVersion calls by the developer principal Correct answer
- ESecurity Health Analytics findings listing secrets that have no rotation period configured
Answer and explanation: Secret Manager Data Access audit logs show which principal read which secret versions, and sign-in events with IPs and user agents reveal whether the account was used from unusual locations or tools, together supporting or refuting the hypothesis. Rotation findings are posture issues, not evidence of access. Billing data is too coarse to attribute access. Subnet flow volume does not show secret access by an identity. Note that Data Access logs must be enabled for Secret Manager to record reads.
Question 5 of 10, Threat Hunting. Choose 2.
Google Threat Intelligence publishes a profile of a threat actor known to target the company's industry, listing common TTPs such as abusing OAuth applications in Google Workspace and creating mailbox forwarding rules. How should the hunt team turn this into an effective hunt? (Choose two.)
- ADerive testable hypotheses for each TTP and identify which Workspace log sources are needed to observe them Correct answer
- BRerun last month's vulnerability scan, since the actor's TTPs are best confirmed through missing patches
- CBlock every OAuth application in Workspace immediately, then close the hunt without reviewing any logs
- DSearch historical Workspace admin and email events for new forwarding rules and unusual OAuth grants Correct answer
- EWait until the actor's indicators appear in curated detections before spending any analyst time hunting
Answer and explanation: Effective intelligence-driven hunting turns TTPs into hypotheses, confirms the telemetry exists, and then searches historical data such as forwarding rule creation and OAuth grant events. Blocking all OAuth apps is a disruptive control decision, not a hunt, and skips evidence gathering. Waiting for curated detections defeats proactive hunting. Vulnerability scans do not reveal abuse of legitimate features like OAuth or mail forwarding.
Question 6 of 10, Detection Engineering
Security wants to detect when a Cloud Storage bucket in a production folder becomes publicly readable, and also when an organization policy that prevents public access is removed from that folder. Detection must work even for changes made outside Terraform. Which approach provides the MOST complete coverage with least custom code?
- AUse SCC Security Health Analytics for public bucket findings and posture drift detection for the deployed organization policy Correct answer
- BCreate a YARA-L rule on bucket object reads from external IPs, because public buckets are only exposed when anonymous users read them
- CWrite a Cloud Function triggered by Cloud Storage object finalize events that checks each new object's ACL for allUsers entries
- DRun a weekly Terraform plan in CI and alert when the plan shows differences for bucket IAM bindings or for organization policies
Answer and explanation: Security Health Analytics detects publicly accessible buckets regardless of how they were changed, and SCC posture management reports drift when a deployed policy such as a public access prevention constraint is altered. A Terraform plan only detects drift periodically and misses non-Terraform resources. Object finalize triggers fire on uploads, not on IAM policy changes. Watching for external reads detects exposure only after data is accessed.
Question 7 of 10, Detection Engineering
A detection engineer wants to flag processes on Windows endpoints that have rarely been seen across the fleet, even though their hashes do not appear in any threat intelligence source. Which Google SecOps capability BEST supports this?
- AUse Applied Threat Intelligence curated detections, which alert on processes whose hashes appear in Mandiant indicators
- BUse Cloud IDS threat logs, which identify rare executables by inspecting mirrored packets in the Google Cloud VPC
- CUse SCC Event Threat Detection, which compares Windows process hashes from endpoints with Google's malware hash lists
- DUse prevalence data from the entity graph in a YARA-L rule or dashboard to find processes seen on very few assets Correct answer
Answer and explanation: Prevalence statistics in Google SecOps show how many assets have seen a process, file, or domain, which is the basis for low-prevalence hunting independent of threat intelligence. Applied Threat Intelligence depends on known indicators, which the scenario rules out. Event Threat Detection analyzes Google Cloud logs rather than endpoint process inventories. Cloud IDS inspects network traffic and does not track process prevalence on endpoints.
Question 8 of 10, Incident Response
Investigators discover that an attacker on a GKE node attempted to access other workloads' identities via the metadata server. The cluster uses Workload Identity Federation for GKE. What is the MOST appropriate assessment and next step?
- ADisable Workload Identity Federation for GKE so pods fall back to the node's service account, which simplifies the forensic analysis
- BAssume all Google Cloud service accounts in the project were compromised and delete each one immediately to stop further damage
- CConclude no impact is possible because Workload Identity Federation for GKE prevents any process on the node from obtaining credentials
- DConfirm the node uses the GKE metadata server so pods get only their own identities, then review which workload identities ran on that node Correct answer
Answer and explanation: With Workload Identity Federation for GKE, the GKE metadata server limits pods to their own identities, but an attacker with node access may reach credentials for workloads on that node, so scoping should review which identities were exposed. Deleting all service accounts without scoping is disruptive. Falling back to the node service account weakens isolation. Node-level compromise can still expose credentials of workloads on that node, so assuming no impact is wrong.
Question 9 of 10, Incident Response
A compromised service account was used to create resources in several projects. The team wants a complete list of actions taken by that principal across the organization over the last 30 days. Which approach is MOST complete?
- ACheck the IAM page of the service account's home project to see which roles it holds in that project today
- BReview the Cloud Billing report for each project to identify resources created during the last 30 days
- CSearch SCC findings for the service account, since every API call it made in any project generated a finding automatically
- DQuery organization-level aggregated audit logs for the principal, covering Admin Activity and enabled Data Access logs Correct answer
Answer and explanation: Organization-level aggregated audit logs capture actions across all projects, including Admin Activity and any enabled Data Access logs, giving the fullest record of the principal's actions. Current IAM roles do not show past actions. Billing reports show costs, not who created resources or what reads occurred. SCC only creates findings for detected threats or misconfigurations, not for every API call.
Question 10 of 10, Observability
A SecOps engineer must be notified on a team chat channel when the volume of logs from any Google SecOps ingestion source drops by more than 80 percent compared with its usual rate. Which Google Cloud setup is MOST appropriate?
- AA YARA-L rule that counts ingested events per hour and sends an email when a source stops matching the rule
- BA Looker Studio report on ingestion volume that the on-call engineer reviews manually every morning at the start of the shift
- CA Cloud Monitoring alerting policy on SecOps ingestion metrics with a threshold condition and a chat notification channel Correct answer
- DA Cloud Scheduler job that restarts all forwarders hourly so that any drop in volume corrects itself automatically
Answer and explanation: Ingestion metrics in Cloud Monitoring support threshold-based alerting policies with notification channels such as chat integrations, providing timely alerts on volume drops. Detection rules are not designed to monitor pipeline health. A manual daily report introduces delay. Restarting forwarders blindly does not detect or explain source-side failures.