Prepare and test your skills
Prepare and test your skills
Data sovereignty, residency, and locality rules decide where regulated data can be stored, where it can be processed, and who can control it. On Google Cloud, architects meet these rules by choosing regional deployment targets and then enforcing them with location constraints, access boundaries, and encryption controls. Cross-border transfer limits and high availability goals directly shape which services and regions remain usable.
Data residency, locality, and sovereignty answer three different compliance questions about the same workload. Data residency describes where Customer Data is stored at rest, such as a specific region or multi-region. Data locality adds where data is processed in use and transmitted in transit, plus where workloads can be accessed from. Data sovereignty adds which jurisdiction governs the data, including which laws apply, which personnel can administer it, and which keys protect it.
| Concept | Core question | Enforced with |
|---|---|---|
| Data residency | Where is data stored at rest? | Resource Locations constraint, regional bucket or dataset location, Assured Workloads folder |
| Data locality | Where is data stored, processed, and accessed? | Regional or zonal services, VPC Service Controls perimeter, Cloud Armor geo rules, Identity-Aware Proxy (IAP) |
| Data sovereignty | Which jurisdiction, personnel, and keys control data? | EU data boundary, routing to EU personnel, access approvals, customer-managed encryption keys |
Residency is the narrowest control, while locality and sovereignty build on it for regulated industries. A workload can satisfy residency but still violate locality if it is processed outside its home region, or violate sovereignty if it is administered from outside its jurisdiction. Architects must therefore check storage location, processing location, access location, and administrative control together.
Enforcing location with organization policies and Assured Workloads prevents resources from being created in the wrong jurisdiction. Organization policies with the Resource Locations constraint define the allowed set of locations for new resources. The constraints/gcp.resourceLocations constraint can be attached at the organization, folder, or project level, and when the policy denies a region, users cannot provision new in-scope resources there.
Assured Workloads packages this enforcement for regulated and sovereign requirements. An Assured Workloads folder applies a selected control package that presets allowed regions and related constraints for every project created inside it. Only resources created inside the compliant folder inherit the protection, so sensitive data with specific residency needs must be placed and kept there. Some resources are out of scope and cannot be restricted, so architects must verify coverage in the supported-services list.
Enforcement follows an ordered dependency:
This order matters because location is fixed early for some services. In Apigee, for example, data residency is selected only during organization provisioning, and the control plane location cannot be changed later without creating a new Google Cloud project.
Designing for locality restrictions and cross-border transfers starts by defining exactly what the regulation limits before choosing services for availability and recovery. Regionality varies by service, and multi-regional resources are resilient to a regional failure while regional resources are resilient only to a zonal failure and zonal resources fail with their zone. If encryption with Cloud Key Management Service (Cloud KMS) (opens in a new tab) can mitigate the rule, multi-regional services can be used with standard disaster recovery techniques. If processing must stay where data rests, architects use regional Compute Engine deployments across zones, product-specific locality settings such as BigQuery dataset locations, or a hybrid approach with Google Distributed Cloud.
Global privacy frameworks add storage and handling rules on top of location. The Health Insurance Portability and Accountability Act (HIPAA) sets United States standards for electronic personal health information (PHI), with Google documenting its infrastructure controls and the customer configuring access controls for PHI on top of them. The General Data Protection Regulation (GDPR) governs customer data for people who reside in the European Union (EU), and the customer must keep that regulated data in Google Cloud regions in the EU. Data flows from users and applications into the selected EU region, processing stays in that region, and supporting artifacts such as inspection jobs and data profiles are also stored in that region.
Access location is controlled separately from storage location. Identity-Aware Proxy verifies user identity before permitting application access, with domain-restricted sharing limiting which domains can appear in Identity and Access Management (IAM) policies. Google Cloud Armor adds IP and geo-based constraints at the edge, while VPC Service Controls provides context-based perimeter security around data and services. For EU workloads, additional sovereign controls include an EU data boundary, support routing to EU persons, visibility into administrative access requests, policy-driven access approvals, and custom encryption options.
Data ownership and shared responsibility define who is accountable for cloud-hosted data and who secures each layer around it. The organization remains the owner of its data in Google Cloud, while Google is responsible for default encryption, infrastructure controls, and enforcing the IAM policies the customer sets. The customer is responsible for classifying data, placing regulated data inside the correct compliant folders or projects, configuring IAM so only appropriate personnel can access it, and organizing the hierarchy so it does not expose personal data.
Establishing ownership for migrated and cloud-generated data means every asset is labeled with an accountable owner and sensitivity level. Data Catalog (opens in a new tab) automatically captures technical metadata from BigQuery (opens in a new tab) tables into a data register, and Tag Engine, an application running on Cloud Run, applies business metadata tags such as owner name and sensitivity level. Sensitive Data Protection inspects assets for infoTypes such as AGE, DATE_OF_BIRTH, PHONE_NUMBER, and EMAIL_ADDRESS, then Tag Engine applies an is_sensitive tag and column-level policy tags. By default, users have no access to columns that have defined policy tags, so a user with table-level IAM access still cannot view a tagged column without the matching policy-tag grant.
Residency commitments and field hygiene affect design. Commitments apply to Customer Data at rest for services that can be configured for data location, and Customer Data does not include resource identifiers, attributes, or other data labels. Service Data such as billing metadata is stored separately. Architects must avoid placing sensitive data in fields outside the technical control boundary, including resource names and IDs, description fields, labels, URIs, file paths, timestamps, user IDs, firewall rules, and customer IAM policies.
Exam tip: Data residency commitments cover Customer Data at rest, not resource names, IDs, labels, or Service Data, so never store health information, personal identifiers, cardholder data, or ID numbers in those common fields.
Commercial sensitive data protection finds confidential cardholder data and personally identifiable information (PII) and keeps it usable without exposing raw values. Sensitive Data Protection (opens in a new tab) automates discovery, classification, and de-identification of credit card numbers, addresses, and phone numbers across storage and analytics services. Discovery shows where sensitive data lives, and de-identification replaces raw values while keeping the data useful for analysis.
Automated discovery and classification continuously scans Google Cloud at the organization, folder, or project level. The discovery service runs on its own, so when a new project is created or a large amount of new data is uploaded, it discovers, classifies, and reports on that data without manual scans. Each scan produces data profiles that include sensitivity level, data risk level, infoType findings such as PHONE_NUMBER or credit card numbers, and location. Data profiles flow to Security Command Center, Google SecOps, BigQuery, and Knowledge Catalog for governance and analysis.
Acting on discovery results depends on whether sensitive data is expected and properly protected. When sensitive data is found where it is expected and controls are correct, profiles should be published to monitoring workflows and kept for audit reports. When sensitive data is found where it was not expected or without proper controls, the workload has high data risk. To lower data risk for BigQuery columns, apply a BigQuery policy tag to restrict access, remove the data if it is not needed, and export profiles to BigQuery before deletion if history must be kept. A column with a high free-text score, especially with multiple infoTypes in unstructured notes, presents added risk because freeform text can hide PII.
De-identification transforms raw values before storage or by creating a de-identified copy, so original PII and cardholder data are masked while analytical value remains:
| Technique | How it protects | Utility preserved |
|---|---|---|
| Masking | Replaces characters or values with redacted symbols | Keeps record structure while hiding the sensitive value |
| Tokenization | Replaces a sensitive value with a token or encrypted value re-identifiable only with proper keys | Keeps format and joinability for processing without exposing raw PII |
| Bucketing | Replaces precise values with broader ranges or categories | Keeps statistical patterns for analysis while reducing identifiability |
Choose masking when raw values must not be visible at all, choose tokenization or format-preserving encryption when downstream systems still need consistent, reversible identifiers under strict control, and choose bucketing when only trends matter. This approach supports data minimization by ensuring teams do not collect or keep more sensitive data than business goals require.
Exam tip: Use discovery for continuous profiling across an organization, folder, or project, and use on-demand inspection when you need the exact locations of PII inside a specific BigQuery table or Cloud Storage bucket.
Access isolation keeps commercial sensitive data inside a tightly controlled boundary. VPC Service Controls (opens in a new tab) is the Layer 7 perimeter service that prevents unauthorized copying of data from Google-managed services. A service perimeter is configured around in-scope projects, and data flows from resources inside the perimeter to the outside only when an egress rule explicitly allows that destination. Requests flow from the outside to the inside only when an ingress rule allows that identity and client. This complements Virtual Private Cloud networks and subnets that provide Layer 3 and Layer 4 isolation, and Cloud Armor that filters IP lists at the HTTP(S) load balancer.
Identity and Access Management (IAM) decides which users, groups, and service accounts can enter the perimeter. A Google Cloud project is a trust boundary and IAM enforcement point, while folders group similar projects and apply folder-level permissions that inherit down to projects. Access follows the principle of least privilege, with IAM Conditions, principal access boundary policies, and Privileged Access Manager for just-in-time elevation. All users who require access to sensitive data must use multifactor authentication as 2-Step Verification, which supports Payment Card Industry Data Security Standard (PCI DSS) requirements for restricted access and authentication.
The cardholder data environment (CDE) is the set of people, processes, and technologies that store, process, or transmit cardholder data. Segmentation combines three layers that build on each other:
| Segmentation layer | What it isolates | How it enforces the boundary |
|---|---|---|
| Logical segmentation | Organization, folder, and project hierarchy | Group all in-scope projects in a dedicated PCI folder and keep in-scope and out-of-scope workloads in different projects |
| Network segmentation | VPC networks and subnets | Place CDE resources in dedicated VPCs and subnets with firewall rules that deny all traffic by default |
| Service-level segmentation | Google-managed APIs and data | Define a VPC Service Controls perimeter around in-scope projects with ingress and egress controls |
Without that isolation, the entire cloud footprint falls in scope, which increases assessment cost and attack surface. A system component counts as out-of-scope only when it is properly isolated so that even if it were compromised, the compromise would not impact the security of the CDE.
Encryption strategy protects sensitive commercial data while proving control over keys to auditors. Google Cloud encrypts customer data stored at rest and data moving in transit by default, with no action required. Regulated workloads use customer-managed keys, hardware protection, and external key storage for stronger proof of control. Cloud Key Management Service (Cloud KMS) (opens in a new tab) provides customer control of key generation, storage, rotation, and destruction for credit card numbers and PII.
Default encryption at rest means Google automatically encrypts stored content using Google-owned keys, which isolates data from lower-level threats such as malicious disk firmware. Envelope encryption adds customer control by wrapping data encryption keys (DEKs) with key encryption keys (KEKs). In the envelope flow, the integrated service encrypts data with a DEK, then Cloud KMS wraps that DEK with a customer-managed KEK, so access to data depends on access to both the encrypted data and the KEK. Customer-managed encryption keys (CMEKs) are created in Cloud KMS and used across Cloud Storage, BigQuery, Cloud SQL, Spanner, Compute Engine disks, Google Kubernetes Engine (GKE) (opens in a new tab) secrets, Pub/Sub, Dataproc, and AlloyDB. When a CMEK is disabled or destroyed, the service can no longer unwrap its DEKs, so the protected data becomes unreadable.
A left-to-right process flow showing plaintext encrypted by a service with a DEK, the DEK wrapped by a customer-managed KEK in Cloud KMS, and stored encrypted data becoming unreadable if the KEK is destroyed.
For regulated workloads, Cloud HSM provides hardware-backed protection inside a Federal Information Processing Standard (FIPS) 140-2 Level 3 certified module, including Single-tenant instances for dedicated isolation. Cloud External Key Manager lets teams store key material outside Google Cloud in a compatible third-party system and link those external keys to Cloud KMS. Encryption in transit protects data moving outside Google-controlled boundaries, with VM-to-VM traffic inside a VPC encrypted, traffic to Google services terminating at the Google Front End, and Application Layer Transport Security securing service-to-service calls. Additional controls include MACsec over Cloud Interconnect, IPsec over Cloud VPN, Transport Layer Security (TLS) with minimum TLS 1.2 for application traffic, and Confidential Computing for data in use in memory.
| Strategy | Key ownership and control | Choose when |
|---|---|---|
| Default encryption at rest | Google-owned keys, no configuration | No specific key-management mandate exists and low overhead is preferred |
| CMEK with Cloud KMS, Cloud HSM, or Cloud EKM | Customer-managed keys with lifecycle, rotation, and external-storage options | Auditors require key management proof, certified hardware, or keys outside Google Cloud |
| Application-layer tokenization before storage | Application-managed tokens plus platform encryption | Highly sensitive fields must not be stored in raw form |
This choice supports PCI DSS Requirements 3 and 4, which require stored primary account numbers (PAN) to be rendered unreadable and PAN in transit over public networks to use strong cryptography with proper key management.
Continuous compliance shows that in-scope components remain secure in practice, not only that the architecture was designed securely. Cloud Logging collects real-time logs and metrics from data pipelines, while Cloud Audit Logs captures audit trails of API calls, including GKE and kubectl requests and Privileged Access Manager elevations. Audit logs support PCI DSS requirements 10 and 11 for logging and testing, but only when teams export, protect, and regularly review them with access controls on the destinations. Export audit logs to Cloud Storage for long-term archival and to BigQuery for analysis and forensics, and use Cloud Monitoring alert policies to notify responders.
Comprehensive coverage for a PCI boundary combines network, perimeter, and access evidence:
Do not write PII or unprotected primary account numbers into experiment logs, email, customer-service logs, or monitoring metadata, because logs can expand PCI scope. Classification therefore determines which projects, datasets, and logs fall inside the protected boundary, a principle documented in HIPAA compliance guidance (opens in a new tab) for handling sensitive identifiers in metadata.
Differentiating certification scopes means choosing the compliance framework whose data types, industries, and security goals match the workload. On Google Cloud, meeting regulatory requirements is a shared responsibility, so the architect must select services whose existing certifications fit the workload requirements. AlloyDB shows how this works because it carries certifications across information security, service controls, payment card, and health domains.
Industry certification groups cover different compliance domains rather than interchangeable approvals. AlloyDB has received certifications in the following groups:
The ISO group addresses information security management and cloud-specific controls, the SOC group addresses operational controls and trust principles, PCI DSS addresses payment card security, and HIPAA addresses health information handling. The compliance resource center is the central source for checking which services hold which certifications. For enforcement, Assured Workloads (opens in a new tab) lets the architect apply controls to specific folders that support regulatory, regional, or sovereign requirements.
Shared responsibility divides audit controls between Google and the customer. Google secures physical infrastructure and network security, applies default encryption, restricts privileged access, and proves its controls through programs such as FedRAMP High Provisional Authority to Operate. The customer is considered the expert on its own data and must weigh regulatory obligations, organizational security standards, and customer and vendor requirements before choosing workload controls. A certification such as SOC 2, ISO 27001, PCI DSS, or FedRAMP is therefore satisfied only when both sides meet their assigned controls.
Exam tip: Security operations center (SOC) analysts who respond to findings are different from Service and Organization Controls (SOC) 1, SOC 2, and SOC 3 certifications.
How responsibility shifts by service model determines how many controls appear in the customer's audit scope:
| Service model | Example services | Responsibility split |
|---|---|---|
| Infrastructure as a Service (IaaS) | Compute Engine, Cloud Storage, Cloud VPN, Cloud Load Balancing, Cloud DNS | The bulk of security responsibility belongs to the customer; Google is responsible for underlying infrastructure |
| Platform as a Service (PaaS) | App Engine, Google Kubernetes Engine (GKE) (opens in a new tab), BigQuery | Google owns more controls; responsibility is shared for application controls and IAM, while the customer remains responsible for data security |
| Software as a Service (SaaS) | Google Workspace, Google Security Operations, Marketplace SaaS | Google owns the bulk of security; the customer remains responsible for access controls and data placed in the application |
In IaaS the architect configures operating systems, networks, and storage access, so misconfiguration risk is highest. In PaaS and SaaS more patching moves to Google, but data classification, identity decisions, and residency choices stay with the architect. Using Assured Workloads and boundaries translates a regime into enforced limits: the architect creates a folder, sets the compliance program such as FedRAMP High or EU controls, and projects under that folder inherit guardrails and region restrictions through the resource-location policy.
Controls the architect must implement include placing regulated data inside appropriate folders and keeping it there, configuring IAM with least-privilege role-based access, organizing the hierarchy so it does not expose personal data, and implementing encryption with Cloud KMS, perimeters with VPC Service Controls, and monitoring with Security Command Center (opens in a new tab). Google enforces the IAM policies the customer sets, but the customer must create them, and information shared during support must avoid moving regulated data outside compliant folders or into names and labels.
Utilizing compliance documentation means collecting third-party reports and technical records that prove a deployment meets requirements. Architects map platform certifications such as ISO 27017 for cloud security, ISO 27018 for personal data protection, PCI-DSS for payment card data, and SOC 1, 2, and 3 for operational controls to customer needs. The Google Cloud Business Associate Agreement also covers VMware Engine (opens in a new tab), which extends protected health information handling to that environment under its shared responsibility model.
Obtaining audit reports and bridge letters means requesting point-in-time proof through the account team rather than generating it inside the project. For assistance with auditing, the architect contacts the account representative for the latest ISO certificates, SOC reports, and self assessments. Those documents provide the proof auditors request, while bridge letters cover the gap period between formal reports. Architects should request the current set during design so controls and timelines match the certified scope.
Using Artifact Registry and supply chain controls provides technical evidence that build artifacts meet security policy. Artifact Registry is a scalable repository for Docker images, Maven packages, and npm packages, with access managed by IAM. Artifact guard evaluates container images against centralized vulnerability policies during build, registry storage, and runtime execution, and it centralizes scan results into a single view in Security Command Center. Policy enforcement follows an ordered workflow where each stage depends on the prior configuration:
Continue with this study guide: Industry certifications (e.g., SOC 2)
Log export and retention compliance keeps audit logs searchable, stored in the right location, and available long enough for forensics and audits. Cloud Logging (opens in a new tab) first stores logs in log buckets in the project where they were generated. Log sinks then route copies to long-term or analytic destinations outside that short-term storage, while retention rules on the buckets control how long the local copy is kept.
Cloud Audit Logs records who did what, where, and when within Google Cloud resources. Logs are written for each project, folder, organization, and billing account, and projects contain only audit logs for resources directly within that project. The four categories divide operations by the kind of action performed:
| Category | Log name suffix | Operations documented |
|---|---|---|
| Admin Activity audit logs | cloudaudit.googleapis.com%2Factivity | Admin write operations that write metadata or configuration |
| Data Access audit logs | cloudaudit.googleapis.com%2Fdata_access | Admin read, data read, and data write operations |
| System Event audit logs | cloudaudit.googleapis.com%2Fsystem_event | System events such as Compute Engine live migration |
| Policy Denied audit logs | cloudaudit.googleapis.com%2Fpolicy | Operations denied by policy |
Admin Activity logs are always enabled and cannot be disabled, so a baseline of configuration changes is always retained. Data Access logs are disabled by default and are not written unless explicitly enabled, because they can be quite large, with the exception of BigQuery Data Access logs which cannot be disabled. Visibility into Google personnel is covered by Access Transparency logs and Access Approval, which lets customers approve access before Google Support accesses data.
An audit log entry is a LogEntry object whose protoPayload field contains an AuditLog object. The AuditLog object holds serviceName that identifies the product that wrote the entry, authentication information, caller IP, timestamp, success or failure status, and affected resource. The part of the log name after /logs/ is URL-encoded, so / appears as %2F. Because every entry names its owner, target resource, time, and writing service, logs can be queried in Logs Explorer by logName, resource.type, and protoPayload.serviceName to trace an action from caller to target.
Exam tip: Admin Activity logs cannot be disabled, while Data Access logs must be explicitly enabled before they are written.
Log buckets are the containers in Cloud Logging that store and organize log data. For every billing account, project, folder, and organization, Cloud Logging automatically creates two buckets, _Required and _Default, with correspondingly named sinks. The _Required bucket stores Admin Activity and System Event logs, and neither its storage nor its sink can be disabled. The _Default bucket stores enabled Data Access and Policy Denied logs by default. Admin Activity logs are still generated even if the Cloud Logging API is disabled.
A log sink is a routing rule that sends some or all logs that match its filter to a destination outside the source bucket. A sink routes a copy, so the original handling in the source bucket still applies unless it is separately changed:
| Destination | What it provides | Choose when |
|---|---|---|
| Cloud Storage | Low-cost object storage for log copies | Long-term archiving is needed for regulatory retention and forensic re-analysis |
| BigQuery | SQL-based dataset for powerful search and analysis | Large volumes of audit logs must be queried quickly across time and projects |
| Pub/Sub | Real-time message delivery to applications | Logs must flow to a SIEM, automation, or external compliance store |
Traffic therefore flows from the source log bucket through the sink filter to Cloud Storage for archiving, to BigQuery for analysis, or to Pub/Sub for real-time forwarding. An aggregated sink created at the organization or folder level routes logs from all projects under it into one central destination, which supports consistent retention and forensic analysis. Admin Activity logs routed to a different project do not pass through the destination project's sinks and are not stored there, so a separate sink must be created in the destination project to store them.
A retention period is the custom rule on a log bucket that defines how long Cloud Logging keeps every log in that bucket. Retention can be configured between 1 day and 3650 days, which allows up to about ten years of local retention. A custom rule applies to all logs in the bucket regardless of type, so separate buckets are needed when different data types require different periods. Data residency is controlled by bucket and destination locations, including default resource locations, while encryption defaults to Google-managed keys with customer-managed keys available for stricter needs. Costs are managed by estimating usage, alerting on usage, and creating exclusions for logs that are not useful, such as excluding Data Access logs from development projects.
Security analysis collects audit and network logs in one place, searches them for unauthorized access, and turns patterns into alerts. Cloud Logging stores the logs, Logs Explorer and Log Analytics provide search and SQL analysis, Cloud Monitoring sends alerts, and Security Command Center correlates findings with log evidence. Together they give visibility into administrative actions and data access across projects, folders, and organizations.
Investigation in Logs Explorer searches audit logs for activity by a compromised identity and for new resources created during the incident window. Access depends on IAM roles: the Logs Viewer role allows viewing audit logs, the Private Logs Viewer role adds Data Access logs with private entries, and the Security Center Findings Viewer role allows listing findings. An investigation follows an ordered flow from logs to findings to containment:
A security log analytics pipeline moves logs from every project into a single destination in three ordered stages. First, enable logs, prioritizing relevant types because Admin Activity is on by default while many Data Access, network, and platform logs must be turned on and incur costs. Second, route logs by creating an aggregated sink at the organization level with a filter that selects the scoped logs. Third, analyze logs with SQL queries in Observability Analytics or BigQuery, YARA-L rules in Google Security Operations, or a third-party SIEM, reusing starting queries from the Community Security Analytics project. Metrics, alerts, and retention then turn stored logs into continuous monitoring, with alerts based on risk for events such as intrusion attempts or firewall violations.
Network and intrusion logs record traffic and attack attempts for near real-time defense. VPC Flow Logs record traffic flows sent from or received by VM instances, with each subnet enabled independently so volume can be reduced by enabling them only on the in-scope environment. Firewall Rules Logging provides per-rule auditing of TCP and UDP connections for rules the architect creates, which verifies whether a deny rule works and warns of unapproved use. Cloud Intrusion Detection System consumes mirrored traffic through Packet Mirroring to gain visibility into north-south and east-west traffic, including VM-to-VM communication that reveals lateral movement.
Continue with this study guide: Audits (including logs)
constraints/gcp.resourceLocations organization policy and Assured Workloads folders prevent creation of location-based resources outside allowed regions.Data residency describes where Customer Data is stored at rest, data locality adds where data is stored, processed in use and transmitted in transit plus where workloads can be accessed from, and data sovereignty adds which jurisdiction governs the data including applicable laws, administering personnel, and protecting keys. A workload can satisfy residency but still violate locality if processed outside its home region or violate sovereignty if administered from outside its jurisdiction, so storage location, processing location, access location, and administrative control must be checked together.
Organization policies with the Resource Locations constraint define the allowed set of locations for new resources and prevent users from provisioning new in-scope resources in a denied region. An Assured Workloads folder applies a selected control package that presets allowed regions and related constraints for every project created inside it, so sensitive data with specific residency needs must be placed and kept inside the compliant folder to inherit protection.
Masking replaces characters or values with redacted symbols when raw values must not be visible at all while keeping record structure. Tokenization replaces a sensitive value with a token or encrypted value re-identifiable only with proper keys when downstream systems still need consistent, reversible identifiers under strict control, while bucketing replaces precise values with broader ranges or categories when only trends matter.
Admin Activity audit logs document admin write operations that write metadata or configuration and are always enabled and cannot be disabled. Data Access audit logs document admin read, data read, and data write operations and are disabled by default and are not written unless explicitly enabled, with the exception of BigQuery Data Access logs which cannot be disabled.
Last updated