Professional Security Operations Engineer (Google Cloud) — Free Practice Questions
10 free sample questions from a bank of 10, with the correct answers and explanations. No signup required — start practising right now.
1You are responsible for identifying suspicious activity and security events at your organization. You have been asked to search in Google Security Operations (SecOps) for network traffic associated with an active HTTP backdoor that runs on TCP port 5555. You want to use the most effective approach to identify traffic originating from the server that is running the backdoor. What should you do?
Detect on events where network.ApplicationProtocol is HTTP.
Detect on events where target.port is 5555.
Detect on events where principal.port is 5555.
Detect on events where network.ip_protocol is TCP.
Answer: C
The short version
C — Detect on events where principal.port is 5555. In UDM, principal is the entity that originates the connection, so this finds the traffic leaving the backdoor server itself.
Key concepts
UDM (Unified Data Model): Google SecOps' normalised schema in which principal is the acting entity and target is the entity being acted upon, with principal.ip/target.ip and their port fields.
principal vs target: for network connection events, principal = source/requester, target = destination/responder; choosing the wrong one flips your direction of travel.
network.ip_protocol / network.ApplicationProtocol: connection-level fields giving the transport protocol (TCP) and the inferred application protocol (HTTP).
Detection rule field choice: the most specific identifier beats a broad protocol filter — port plus direction is far more selective than "is this TCP?".
Why C is correct
The question asks for traffic originating from the server running the backdoor. UDM models the initiator of an action as the principal, so the backdoor server is the principal in its outbound sessions and its listening/service port is recorded in principal.port. Filtering on principal.port is 5555 therefore returns exactly the sessions emanating from the backdoor host on that port. It is also the most selective predicate offered: it pins both the direction of travel and the port, which is what lets an analyst pivot to the specific host rather than wade through every TCP or HTTP connection in the tenant.
Why the others are wrong
A.network.ApplicationProtocol is HTTP matches essentially all web traffic tenant-wide. It is the broadest possible filter, says nothing about the port and cannot isolate the backdoor's traffic.
B.target.port is 5555 selects the receiving side of connections to port 5555, i.e. sessions where the backdoor host is the destination. That is the wrong direction for traffic originating from the server.
D.network.ip_protocol is TCP matches all TCP sessions. It is a protocol filter with neither port nor direction, so it produces enormous noise and never narrows to the backdoor.
PSOE exam tip
Direction first: UDM principal = who initiated, target = who received. Then narrow by port. Protocol-only filters are never the answer to a "which host" question.
2You are an incident responder at your organization using Google Security Operations (SecOps) for monitonng and investigation. You discover that a critical production server, which handles financial transactions, shows signs of unauthorized file changes and network scanning from a suspicious IP address. You suspect that persistence mechanisms may have been installed. You need to use Google SecOps to immediately contain the threat while ensuring that forensic data remains available for investigation. What should you do first?
Use the firewall integration to submit the IP address to a network block list to inhibit internet access from that machine.
Deploy emergency patches, and reboot the server to remove malicious persistence.
Use the EDR integration to quarantine the compromised asset.
Use VirusTotal to enrich the IP address and retrieve the domain. Add the domain to the proxy block list.
Answer: C
The short version
C — Use the EDR integration to quarantine the compromised asset. Quarantine isolates the host at the endpoint layer, stops the malware's execution and outbound traffic, and leaves the image and logs intact for forensics.
Key concepts
Containment: isolating an asset so the attacker cannot spread further, while deliberately avoiding steps that destroy evidence or change system state before scoping is complete.
EDR quarantine: an endpoint action that isolates the device (network containment and/or process/file quarantine) via the EDR integration, keeping collected telemetry and the ability to acquire a forensic image.
Order of operations: contain → preserve evidence → investigate/scope → eradicate → recover; containment and evidence preservation come first, eradication only after scoping.
Block lists vs asset isolation: firewall/proxy block lists stop traffic to a known-bad indicator, but they leave the infected host running and running malware.
Why C is correct
The requirement is immediate containment while preserving forensic data. The EDR integration's quarantine action is built for exactly that trade-off: it cuts the host off at the endpoint, terminates or isolates the malicious process, and stops further movement, yet it leaves the system in a state that can still be imaged and whose telemetry remains in Google SecOps. That satisfies both halves of the ask. The firewall block-list option targets an indicator rather than the asset, and the other choices either destroy volatile evidence or fail to contain anything at all.
Why the others are wrong
A. Submitting the IP to a network block list stops traffic to and from that IP; the infected production server keeps executing malware, keeps writing files and can pivot to new addresses. It also is a later indicator-driven step, not the first containment move.
B. Emergency patching plus a reboot is an eradication action taken before scoping. A reboot wipes volatile memory, running processes and network state — exactly the forensic data the scenario says must be preserved — and may fail to remove persistence anyway.
D. VirusTotal enrichment is an investigation/enrichment step, not containment, and adding the domain to the proxy block list affects every user behind the proxy, creating collateral outage. Nothing isolates the compromised server.
PSOE exam tip
First move in an incident: contain the asset with EDR, not the indicator with a block list. Never patch or reboot before you have preserved evidence.
3Your organization uses Google Security Operations (SecOps). You discover frequent file downloads from a shared workspace within a short time window. You need to configure a rule in Google SecOps that identifies these suspicious events and assigns higher risk scores to repeated anomalies. What should you do?
Configure a rule that flags file download events with the highest risk score, regardless of time frame.
Create a frequency-based YARA-L detection rule that assigns a risk outcome score and is triggered when multiple suspicious downloads occur within a defined time frame.
Configure a single-event YARA-L detection rule that assigns a risk outcome score and is triggered when a user downloads a large number of files in 24 hours.
Enable default curated detections, and use automatic alerting for single file download events.
Answer: B
The short version
B — Create a frequency-based YARA-L detection rule that assigns a risk score and fires when multiple suspicious downloads occur inside a defined time window. Repetition-in-time is exactly what a frequency-based rule measures.
Key concepts
YARA-L 2.0 rule sections: meta, event (matches one UDM event), match (aggregates those events), and condition (when to fire, and with what outcome/risk).
Frequency/threshold detection: uses match variables such as count of matching events, distinct users or destinations within a time frame — the right pattern for bursty behaviour.
Risk score / outcome: assigning a numeric outcome (e.g. 100) in the match section so repeated occurrences accumulate risk and drive case prioritisation.
Single-event rule: fires on each qualifying event, so it cannot express "N times within T minutes" and gives every event identical weight.
Why B is correct
The scenario has two requirements: identify the burst of downloads, and assign higher risk to repeated anomalies. Only a rule that counts occurrences within a time window can do both. In YARA-L, that means a frequency-based rule: the event section selects the download events, the match section applies a count/over-time threshold and sets a risk outcome, and the condition section fires only when the threshold is crossed. That converts a stream of individually ambiguous downloads into a high-priority, scored detection — which is precisely what the stem asks for.
Why the others are wrong
A. Assigning the highest risk score to every download event regardless of timeframe removes the context that makes the activity suspicious. It produces maximum noise and does not measure repetition at all.
C. This describes a single-event rule, which cannot count multiple events; the "large number of files in 24 hours" wording also collapses repetition into one event's size, so escalating risk with recurrence is lost.
D. Enabling curated detections plus automatic alerting on single download events adds no frequency logic and no risk scoring — every download alerts with equal weight, which is the noise problem in the first place.
PSOE exam tip
Ask "is the signal in one event or in many events over time?" Many-events-in-a-window means frequency/threshold rule with a risk outcome; one event means a plain event rule.
4You are implementing Google Security Operations (SecOps) at your organization. You discover that the current detection rules are too noisy. Due to the high volume of alerts, some true positives might be missed. You want to ingest additional context sources to reduce false positives in your security detections and to improve the overall positive ratio of the alerts. What should you do?
Ingest high-value asset (HVA) data from your configuration management database (CMDB) system to increase the priority of the alerts based on the sensitivity of the assets found in the detection rules.
Ingest dark web forum handlers from your threat intelligence system to match dark web principals within the detection rules.
Ingest IOCs from your threat intelligence system to validate the IP addresses, domains and hashes with the detection rules.
Ingest tactics, techniques, and procedures (TTPs) from your threat intelligence system to validate the processes and tools with the detection rules.
Answer: A
The short version
A — Ingest high-value asset data from the CMDB so detections can be prioritised by the sensitivity of the asset they hit. Asset criticality context is what separates a real incident from a noisy hit on a test box.
Key concepts
Entity data / context enrichment: a data source that maps IPs, usernames and devices to business attributes, so rules can reason about where an alert landed.
High-value assets (HVA): business-critical systems such as finance, identity and production infrastructure, ranked by criticality from the CMDB.
Asset criticality in UDM: target.asset_criticality / principal.asset_criticality values (low/medium/high) that rules can match or use as outcome thresholds.
Positive ratio: the share of alerts that are genuinely actionable; it improves when context lets you suppress or de-prioritise low-consequence activity.
Why A is correct
The stated problem is noise drowning out true positives. The cheapest, most reliable fix is context: if every rule knows the business criticality of the affected asset, you can raise the outcome or require a match on high-criticality assets and stop treating a scan against a lab VM the same as activity on the finance server. Ingesting HVA/entity data from the CMDB populates exactly those asset fields, which is why A is the right ingestion. The threat-intelligence options add indicators about who the attacker is, not about what was hit — they do not correct the volume problem the scenario describes.
Why the others are wrong
B. Dark web forum "handles" identify adversary personas, not assets. They carry almost no signal for this tenant and cannot tell the SOC whether an alert touches critical infrastructure.
C. IOC ingestion validates IP/domain/hash indicators against known-bad lists. That filters alerts on known threats, but here the threat is unknown and the issue is sheer volume — adding IOCs adds data, not asset context.
D. TTPs are abstract behavioural patterns; they rarely match directly in a detection rule and, again, describe adversary behaviour rather than the importance of the impacted asset.
PSOE exam tip
Noise fix in one question: are you alerting on too many things, or on unimportant things? Unimportant → ingest asset/HVA context (CMDB); unknown threat → ingest IOCs.
5You are developing a new detection rule in Google Security Operations (SecOps). You are defining the YARA-L logic that includes complex event, match, and condition sections. You need to develop and test the rule to ensure that the detections are accurate before the rule is migrated to production. You want to minimize impact to production processes. What should you do?
Develop the rule logic in the UDM search, review the search output to inform changes to filters and logic, and copy the rule into the Rules Editor.
Use Gemini in Google SecOps to develop the rule by providing a description of the parameters and conditions, and transfer the rule into the Rules Editor.
Develop the rule in the Rules Editor, define the sections the rule logic, and test the rule using the test rule feature.
Develop the rule in the Rules Editor, define the sections of the rule logic, and test the rule by setting it to live but not alerting. Run a YARA-L retrohunt from the rules dashboard.
Answer: C
The short version
C — Build the rule in the Rules Editor, define the event/match/condition sections, and validate it with the test rule feature. Testing happens inside the Rules Editor against sample data, so nothing touches the production alerting pipeline.
Key concepts
Rules Editor: the dedicated authoring surface for Google SecOps rules, with version control, staged rollout and built-in validation of the rule structure.
Test rule: runs the rule against a selected sample of data and reports matched events and outcomes without deploying the rule or generating alerts.
Rule lifecycle: draft → test (Rules Editor) → enable in live mode → alert → later retrohunt for historical data.
Retrohunt: re-runs an existing rule over a past time range on stored data; it is a hunting tool, not a rule-validation tool.
Why C is correct
The requirement is to validate complex rule logic before production, with minimal production impact. The Rules Editor plus its test feature is precisely the offline path: you author the YARA-L sections, point the test at a representative data sample, inspect matched events and outcomes, iterate until the logic is right, and only then enable the rule. Nothing is deployed, no alerts are generated and no live traffic path is altered while you are validating, which is exactly the "minimise impact to production" constraint.
Why the others are wrong
A. Exploring in UDM search is useful for learning the data shape, but a search result is not a rule — nothing is versioned, tested or deployable, and you still end up copying unvalidated logic into the editor.
B. Gemini in Google SecOps can suggest rule logic from a natural-language description, but it does not test or validate the rule against data. Nothing verifies accuracy before production.
D. Setting the rule live (even without alerting) still runs it through the production rule pipeline over live data, and a retrohunt re-runs it across historical data at full scale — that is production-scale processing, not isolated validation.
PSOE exam tip
Learning happens in UDM search; authoring and validating happen in the Rules Editor with test rule; hunting the past happens with retrohunt. Three tools, three jobs.
6Your organization has recently acquired Company A, which has its own SOC and security tooling. You have already configured ingestion of Company A's security telemetry and migrated their detection rules to Google Security Operations (SecOps). You now need to enable Company A's analysts to work their cases in Google SecOps. You need to ensure that Company A's analysts:
do not have access to any case data originating from outside of Company A.
are able to re-purpose playbooks previously developed by your organization's employees.
You need to minimize effort to implement your solution. What is the first step you should take?
Acquire a second Google SecOps SOAR tenant for Company A.
Provision a new service account for Company A.
Define a new SOC role for Company A.
Create a Google SecOps SOAR environment for Company A.
Answer: D
The short version
D — Create a Google SecOps SOAR environment for Company A. A separate nested environment gives Company A its own isolated case data while still allowing playbooks to be re-used across environments.
Key concepts
SOAR environment: an organisational boundary inside a single Google SecOps instance, with its own data scope, views, playbooks and case handling, nested under the parent environment.
Case data isolation: users of a child environment can only see cases and data belonging to that environment, which satisfies the "no access to external case data" requirement.
Playbook re-use: playbooks can be copied between environments and customised, so Company A analysts benefit from work already done without a second platform.
Multi-tenancy vs multi-instance: separate tenants duplicate licensing, ingestion and maintenance effort; nested environments achieve the boundary with far less work.
Why D is correct
Both requirements point at the same structural answer. First, Company A's analysts must not see case data from outside Company A — that requires a data boundary, which a nested SOAR environment provides through its own data scope and permissions. Second, they must be able to re-purpose playbooks the parent organisation already built — nested environments are designed for exactly that sharing pattern. Because one requirement implies an environment anyway, the same step satisfies the second, minimising effort. A second tenant would satisfy isolation but at the cost of duplicating ingestion, licensing and administration.
Why the others are wrong
A. A second tenant gives strong isolation but is the maximum-effort option: duplicate licences, duplicate ingestion setup and duplicate rule/playbook maintenance for the same outcome an environment boundary achieves.
B. A service account is a machine identity for integrations and API automation. It creates no boundary, so Company A's analysts would still see the full data set in the shared instance.
C. A new SOC role grants permissions inside the existing environment. Roles control who can do what, not which data they can see — it cannot deliver the data-isolation requirement on its own.
PSOE exam tip
"Isolate company data but share the work?" → nested SOAR environment. Roles change privileges, environments change data visibility; only the environment isolates.
7You have identified and isolated a new malware sample installed by an advanced threat group that you believe was developed specifically for an attack against your organization. You want to quickly and efficiently analyze this malware to get IOCs without alerting the threat group. What should you do?
Search for the threat group in Google Threat Intelligence.
Upload the malware to Google Threat Intelligence by using VirusTotal.
Upload the malware to Google Threat Intelligence by using Private Scanning.
Calculate the file checksum for the malware, and search for the checksum in GoogleThreat Intelligence by using VirusTotal.
Answer: C
The short version
C — Upload the malware to Google Threat Intelligence using Private Scanning. Private Scanning analyses the sample privately inside Google Threat Intelligence, producing sandbox and extraction IOCs without notifying the threat group.
Key concepts
Private Scanning: a Google Threat Intelligence / VirusTotal capability that submits a file to a private, exclusive scanning environment and returns sandbox detonation results, strings and IOCs — visible only to your organisation.
Community sharing (standard VirusTotal): uploading to the public VirusTotal service shares the sample and hash with the security community, where a targeted actor can discover the submission.
Sandbox detonation: executing the sample in an instrumented environment to observe behaviour, dropped files, registry changes, domains contacted and dropped payloads — the usual source of new IOCs.
Hash lookup vs submission: searching an existing hash retrieves prior reports only; a bespoke, never-seen sample yields nothing until it is analysed.
Why C is correct
Two constraints drive this: you need IOCs quickly from actual analysis, and you must not tip off a group that has specifically targeted you. Private Scanning satisfies both — detonation and static analysis happen in Google's sandbox, and the sample is never published to the community feed, so the adversary cannot see that their malware has been submitted. For a bespoke sample this is the only listed path that actually produces new IOCs quietly.
Why the others are wrong
A. Searching the threat group's name in Google Threat Intelligence finds their known campaigns and existing reports. It tells you about the group but does not analyse this sample, so it produces no sample-derived IOCs.
B. Uploading to VirusTotal through the standard service exposes the file to the community and alerts anyone monitoring the service — precisely the disclosure the scenario forbids, and it can delay analysis pending other submissions.
D. Hashing and searching only returns what others already reported about that exact hash. A newly built, never-submitted sample returns no analysis, so you get no IOCs out of it.
PSOE exam tip
Targeted malware + need for IOCs + do not tip off = Private Scanning. Anything that touches the public VirusTotal feed is a disclosure risk.
8Your organization uses Cloud Identity as their identity provider (IdP) and is a Google Security Operations (SecOps) customer You need to grant a group of users access to the Google SecOps instance with read-only access to all resources, including detection engine rules. How should this be configured?
Create a Google Group and add the required users. Grant the roles/chronicle.Viewer IAM role to the group on the project associated with your Google SecOps Instance.
Create a Google Group and add the required users. Grant the roles/chronicle.limitedViewer IAM role to the group on the project associated with your Google SecOps instance.
Create a workforce identity pool at the organization level. Grant the roles/chronicle.editor IAM role to the principalSet://iam.googleapis.com/locations/global/workforcePools/POOL_ID/group/GROUP_ID principal set on the project associated with your Google SecOps instance.
Create a workforce identity pool at the organization level Grant the roles/chronicle.limitedViewer IAM role to the principalSet://iam.googleapis.com/locations/global/workforcePools/POOL_ID/group/GROUP_ID principal set on the project associated with your Google SecOps Instance.
Answer: A
The short version
A — Create a Google Group with the users and grant roles/chronicle.viewer on the project hosting the Google SecOps instance. Viewer is the read-only predefined role that includes detection engine rule visibility.
Key concepts
Predefined IAM roles: Google Cloud's bundled roles such as roles/chronicle.viewer (read-only across Google SecOps resources) and roles/chronicle.limitedViewer (a narrower read-only role without full detection rule access).
Group-based access: binding IAM to a group rather than individual users, so onboarding and offboarding are managed centrally.
Workforce identity pool: federated identity for external/partner/workforce users via principal sets; unnecessary when Cloud Identity is already the IdP.
Editor vs viewer: editor grants write access to detections, playbooks and configuration — broader than read-only access.
Why A is correct
The requirement is read-only access to all resources including detection engine rules, for users whose IdP is Cloud Identity. The predefined Viewer role on the SecOps project is exactly read-only across the instance and includes the detections/rules content, and binding it to a Google Group is the cleanest administration path for a cohort of Cloud Identity users. The narrow roles and the federated-pool options fail on either the "all resources / rules" requirement or on identity source.
Why the others are wrong
B.roles/chronicle.limitedViewer is deliberately restricted and does not grant the full read-only access across all resources, including detection engine rules, that the scenario requires.
C. A workforce identity pool plus principalSet is the mechanism for federated external/workforce identities. With Cloud Identity as the IdP it adds needless complexity, and chronicle.editor grants write access, not read-only.
D. Same identity-source problem as C, and limitedViewer again falls short of read-only access to everything including detection rules.
PSOE exam tip
Google SecOps access control: pick the predefined role first (viewer = full read-only), then pick the identity source (group for Cloud Identity, workforce pool for federated external users).
9Your team is responsible for cybersecurity for a large multinational corporation. You have been tasked with identifying unknown command and control nodes (C2s) that are potentially active in your organization's environment. You need to generate a list of potential matches within the next 24 hours. What should you do?
Write a rule in Google Security Operations (SecOps) that scans historic network outbound connections against ingested threat intelligence Run the rule in a retrohunt against the full tenant.
Load network records into BigQuery to identify endpoints that are communicating with domains outside three standard deviations of normal.
Review Security Health Analytics (SHA) findings in Security Command Center (SCC).
Write a YARA-L rule in Google Security Operations (SecOps) that compares network traffic of endpoints to low prevalence domains against recent WHOIS registrations.
Answer: D
The short version
D — Write a YARA-L rule that compares endpoint network traffic against low-prevalence domains with recent WHOIS registrations. Newly registered, rarely-visited domains are the classic signature of unknown command-and-control infrastructure.
Key concepts
Unknown C2 hunting: identifying attacker infrastructure that no IOC list covers, using behavioural and registration-age signals rather than known-bad indicators.
Low-prevalence domain: a domain very few endpoints in the enterprise have ever resolved or connected to — an outlier in the DNS/connect telemetry.
WHOIS registration recency: registration date as a proxy for infrastructure freshness; adversary domains are typically days or weeks old.
YARA-L rule sections: the event section selects DNS/network events, match correlates low-prevalence destinations with recent registrations, condition sets the threshold and outcome.
Why D is correct
The task is to surface unknown C2 nodes quickly, which rules out anything depending on curated indicator lists or on long retrospective analysis. D builds the detection from the enterprise's own traffic plus public registration data: the event section picks outbound connections to low-prevalence domains, the match section correlates them with recent WHOIS registrations, and the condition sets the threshold. Newly registered domains with almost no enterprise history are precisely the profile of freshly built C2 nodes, and a live rule produces the candidate list within hours rather than days.
Why the others are wrong
A. A retrohunt over the full tenant re-processes large volumes of historical data and, being a hunt rather than a live rule, cannot deliver a list within 24 hours. Scanning against ingested threat intelligence only finds known C2, not unknown nodes.
B. Exporting network records into BigQuery and manually analysing deviation is an ad hoc, engineering-heavy exercise with no detection logic and no recurring coverage — it will not produce a list in a day.
C. Security Health Analytics findings in Security Command Center surface misconfigurations and posture issues for Google Cloud resources. They do not hunt for unknown external C2 domains.
PSOE exam tip
Unknown C2 = freshness plus rarity. Domains that are new (WHOIS) and rare (low prevalence in your own traffic) are the pair to alert on; IOC feeds only catch the known.
10You received an alert from Container Threat Detection that an added binary has been executed in a business critical workload. You need to investigate and respond to this incident. What should you do? (Choose two.)
Notify the workload owner. Follow the response playbook, and ask the threat hunting team to identify the root cause of the incident.
Review the finding, investigate the pod and related resources, and research the related attack and response methods.
Review the finding, quarantine the cluster containing the running pod, and delete the running pod to prevent further compromise.
Silence the alert in the Security Command Center (SCC) console, as the alert is a low severity finding.
Keep the cluster and pod running, and investigate the behavior to determine whether the activity is malicious.
Answer: A, B
The short version
A and B — Notify the workload owner and run the response playbook with threat hunting for root cause, while reviewing the finding and investigating the pod and related resources. Containment and evidence preservation come before any destructive action.
Key concepts
Container Threat Detection: a Security Command Center service that analyses GKE audit and runtime telemetry to surface suspicious behaviour in containers, such as an unexpected binary being executed in a pod.
Pod/container forensics: inspecting the workload, its image, referenced secrets, service account and network peers before anything is deleted; container and node logs expire quickly.
Response playbook: the predefined SOAR/response workflow that drives notification, scoping, containment and recovery consistently.
Evidence preservation and scope: isolating or deleting a running workload destroys the volatile artifacts (process tree, filesystem, memory-backed state) needed to determine the initial access vector.
Why A and B are correct
Both selections do the work an incident responder must do first.
B. is the investigative core: review the Security Command Center finding, then examine the pod, the image and the related resources — nodes, service account, secrets, network peers — and research the technique involved, because container evidence is short-lived and must be gathered before anything changes.
A. is the escalation core: inform the accountable workload owner, then follow the response playbook and hand root-cause determination to threat hunting.
Together they scope the incident while preserving the artifacts needed to explain it, and both can run in parallel.
Why the others are wrong
C. Quarantining the whole cluster and deleting the running pod destroys the compromised container's filesystem, process state and network history — precisely the evidence required to determine how the binary got there. Blast radius also exceeds the incident.
D. Silencing an alert is an alert-hygiene action, not a response. Nothing has been scoped, and an unexplained binary execution in a business-critical workload is not a finding to mute.
E. Leaving a suspected malicious workload running preserves volatile evidence, but it also leaves the adversary executing in a critical system. With a Security Command Center alert indicating an added binary was executed, the correct move is parallel investigation and containment through the playbook, not indefinite observation.
PSOE exam tip
Container incident order: investigate the pod and preserve evidence, notify the owner, run the playbook, contain last. Never delete the workload before you know the initial access.