PCSE question patterns
Find a pattern
Each row is wording that keeps showing up in questions, with the answer it usually points to in green. Tap a row to see why.
In this list
No pattern matches that. Try a service name, such as CodeDeploy or Config.
Wording that decides the answer
-
An organization policy constraint (boolean, list, managed or custom) set at the organization or folder.
Org policy blocks the setting at creation, regardless of who holds IAM roles.
-
An IAM deny policy on the permission, with exceptions for a break-glass group.
Deny is evaluated before allow and cannot be bypassed by inherited roles.
-
The managed feature: CMEK Autokey, Security Command Center, Cloud Armor preconfigured rules, Audit Manager.
Custom scripts and manual reviews carry maintenance the question is trying to avoid.
Identity and federation
-
Google Cloud Directory Sync for users and groups plus SAML SSO to the identity provider.
GCDS provisions accounts one way, SSO authenticates. Neither does both jobs.
-
Workforce Identity Federation with an organization-level pool and attribute conditions.
External identities are mapped to principals without Cloud Identity accounts.
-
Workload Identity Federation with a provider, attribute condition on repo or account, and either direct access or service account impersonation.
The external token is exchanged for a short-lived Google token.
-
Few dedicated accounts, security keys enforced, not used day to day, sign-in alerts, break-glass procedure.
They have unlimited tenant control, so phishing-resistant 2SV and minimal use are the controls.
Service accounts and authorization
-
Grant Token Creator on that service account and use impersonation, not a downloaded key.
Impersonation gives short-lived tokens and logs the developer.
-
Enforce iam.disableServiceAccountKeyCreation at the organization.
It is a preventive constraint, applied to every project.
-
Policy Troubleshooter for one principal-permission pair, Policy Analyzer for who-can-access queries.
Troubleshooter evaluates a specific request. Analyzer queries the policy graph.
-
Privileged Access Manager entitlement with justification, approver and maximum duration.
It issues a temporary role binding and audits it, so no standing privilege exists.
-
Access Context Manager access levels used in IAM conditions, IAP or VPC Service Controls.
Access levels express device, location and identity context.
-
Organization policy iam.allowedPolicyMemberDomains.
It blocks granting roles to outside or consumer accounts and allUsers-style exposure.
Network protection
-
Cloud Armor policy with preconfigured WAF rules, rate limiting, and Adaptive Protection on the external Application Load Balancer.
Cloud Armor filters at Google's edge before traffic reaches backends.
-
IAP TCP forwarding with a firewall rule allowing 35.235.240.0/20.
IAP authenticates by identity and tunnels without public exposure.
-
VPC Service Controls perimeter, tested in dry-run first, with ingress and egress rules for legitimate flows.
IAM allows the user, only the perimeter blocks the API call crossing the boundary.
-
Cloud NGFW Enterprise with firewall endpoint, security profile group and TLS inspection using a CA Service pool.
Only the Enterprise tier does layer 7 prevention and TLS inspection.
-
Secure Web Proxy with URL list rules.
Cloud NAT does no filtering, and FQDN firewall rules lack URL-path control.
-
Target service accounts or IAM-controlled secure tags, not network tags.
Network tags can be set by VM editors, so they are not a trust boundary.
-
VPN or Interconnect, DNS mapping to restricted.googleapis.com (199.36.153.4/30), or Private Service Connect for Google APIs.
The restricted VIP exposes only APIs that a perimeter can control.
-
HA VPN, or Cloud Interconnect with MACsec or HA VPN over Interconnect.
Interconnect is private but not encrypted by default.
-
Private Service Connect endpoint to the service attachment.
It exposes only the service and avoids overlapping IP problems.
Data protection and keys
-
Sensitive Data Protection format-preserving encryption with a KMS-wrapped key.
FPE keeps length and character set and is reversible with the key.
-
Cryptographic hash (HMAC) de-identification.
Hashing is stable for joins and irreversible.
-
Sensitive Data Protection discovery (data profiles) at organization or folder level.
It profiles continuously without per-table jobs.
-
Policy tags from a Data Catalog taxonomy, with Fine-Grained Reader, or dynamic data masking.
Column-level security is the BigQuery native way. Row access policies filter rows.
-
CMEK in Cloud KMS, revoke by disabling or destroying the key.
Google default keys can't be revoked by the customer.
-
Cloud External Key Manager, with Key Access Justifications if reasons must be checked.
The external key manager holds the key, and Google calls it for every operation.
-
Cloud KMS Autokey plus gcp.restrictNonCmekServices.
Autokey provisions keys on resource creation, and the policy forces CMEK use.
-
Secret Manager with secretAccessor granted to the workload service account on that secret, mounted rather than baked in.
It gives versioning, rotation notifications and audit, unlike env vars in code.
-
Confidential VMs, Confidential GKE Nodes or Confidential Space.
Memory is hardware-encrypted. Shielded VMs only protect boot integrity.
-
Model Armor templates, with Sensitive Data Protection for data leakage.
It's the dedicated AI request and response filter.
Operations and detection
-
Binary Authorization policy requiring attestations from KMS-keyed attestors, rolled out in dry-run first.
It enforces at deploy time and supports separation of duties across attestors.
-
Artifact Analysis on-demand scanning in a Cloud Build step.
It scans a local image prior to push and gates the build.
-
VM Manager OS patch deployments and vulnerability reports.
It's managed, scheduled and reports per instance.
-
Data Access audit logs, provided they were enabled beforehand.
Admin Activity logs only show configuration changes.
-
Organization aggregated sink to a locked log bucket or a Cloud Storage bucket with retention lock.
An org sink can't be bypassed per project and locks stop deletion.
-
Grant the sink's writer identity the right role on the destination.
Sinks write with their own service account, which needs permission.
-
Log sink to Pub/Sub, then a Dataflow template (or a native integration).
Pub/Sub streams logs near real time to external consumers.
-
Packet Mirroring to a collector behind an internal load balancer.
Flow logs contain no payload.
-
Cloud IDS endpoint with a packet mirroring policy.
It detects threats and logs them, and doesn't block.
-
Security Command Center Premium, Security Health Analytics, with a custom module for company-specific rules.
SHA detects after the fact. Custom org constraints prevent creation.
-
SCC notification to Pub/Sub triggering Cloud Run or Workflows that remediates, with approval for risky steps.
Event-driven response needs no polling.
-
Isolate with a deny rule on its tag or service account, snapshot disks to a forensics project, preserve logs, then rebuild from a clean image.
Contain first and keep evidence before destroying the instance.
Compliance
-
Assured Workloads folder with the matching compliance regime.
It applies location, service, key and personnel controls and monitors drift.
-
Access Approval, which needs Access Transparency enabled.
Transparency logs access, and Approval lets you allow or dismiss it first.
-
gcp.resourceLocations with in:eu-locations, regional resources, same-region keys, logs and backups.
The policy blocks creation elsewhere, but only for new, supported resources.
-
Audit Manager, with SCC compliance views for continuous status.
It assesses controls and exports evidence for auditors.
-
Isolate the cardholder environment in dedicated projects and VPC, restrict peering and access, add a VPC-SC perimeter.
Only segmented systems stay out of scope.
Now spot them in real questions. Practice sets use the same wording.
Practise 20 questions