Prepare and test your skills
Prepare and test your skills
Automated discovery and de-identification of sensitive data is the process of finding confidential commercial data and transforming it so it can be used safely. Sensitive Data Protection (opens in a new tab) automates discovery, classification, and protection of credit card numbers, addresses, phone numbers, and other personally identifiable information (PII). Discovery shows where sensitive data lives across storage and analytics services, and de-identification replaces raw values while keeping the data useful for analysis.
Automated discovery and classification in Sensitive Data Protection 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 from discovery flow to governance and analysis destinations, including Security Command Center (opens in a new tab), Google SecOps, BigQuery, and Knowledge Catalog (Dataplex Catalog). A premade reporting view breaks down findings by risk, by infoType, and by location, which helps teams see where low-sensitivity and high-sensitivity data exist across countries and projects. Teams first define what counts as confidential, then use these classifications to tag data and decide on protection before it is stored or shared.
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.
Acting on discovery results depends on whether sensitive data is expected and properly protected in the workload where it was found. When sensitive data is found where it is expected and controls are already correct, no remediation is required, but profiles should be published to monitoring workflows and kept for audit reports. Publishing profiles to Security Command Center or Google SecOps helps determine the severity of a threat to sensitive data, while publishing to Knowledge Catalog tracks profile metrics alongside business metadata.
When sensitive data is found where it was not expected or without proper controls, the workload has high data risk because sensitive information exists without additional protection. To lower data risk for BigQuery columns that contain sensitive data, apply a BigQuery policy tag to restrict access to only accounts with specific rights. If the high-risk data is not needed, remove it, and if history must be kept for auditing, configure the profiler to export data profiles to BigQuery before deletion. A column with a high free-text score, especially one with multiple infoTypes in unstructured notes or comment fields, presents added risk because freeform text can hide PII such as birth dates in sentences.
De-identification in Sensitive Data Protection transforms raw sensitive values so analysts and engineers can work without seeing raw identifiers. The transformation is applied before storage or by creating a de-identified copy, so the original PII and cardholder data are masked or replaced while analytical value remains. For Cloud Storage data, the service can make de-identified copies directly, and for broader pipelines teams can ingest, tokenize, and then persist only tokenized data to destinations such as BigQuery. This approach supports data minimization by ensuring teams do not collect, keep, or use more sensitive data than business goals require.
Common transformation techniques split by how they hide the original value and what utility they preserve:
| Technique | How it protects | Utility preserved |
|---|---|---|
| Masking | Replaces characters or values with redacted symbols | Keeps record structure while hiding the sensitive value |
| Tokenization (pseudonymization, including format-preserving encryption) | Replaces a sensitive value with a token or encrypted value that can be re-identified only with proper keys | Keeps format and joinability for processing without exposing raw PII or card numbers |
| Bucketing (generalization) | 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. De-identification works best with layered controls such as policy tags, access restrictions, and avoiding PII in logs, experiment outputs, and metadata.
Access isolation and regulatory auditing keeps commercial sensitive data inside a tightly controlled boundary and proves with logs that the boundary holds. VPC Service Controls (opens in a new tab) creates the data perimeter that blocks exfiltration, Identity and Access Management (IAM) (opens in a new tab) controls who can enter it, and Cloud Logging (opens in a new tab) records what happened for PCI DSS and other commercial frameworks. Together they reduce PCI scope to only the systems that must handle cardholder data and provide continuous evidence of compliance.
Classifying sensitive commercial data is the required first step before that data can be isolated and protected. Sensitive commercial data includes confidential cardholder data and PII that must be identified across stores such as Cloud Storage buckets, BigQuery datasets, Cloud SQL, Datastore, and Spanner. Sensitive Data Protection discovers and classifies this data and can then tag, mask, truncate, or tokenize values before storage.
Knowledge Catalog (formerly Data Catalog) provides the central platform for storing, managing, and accessing that classification metadata, including data lineage, quality profiles, and inspection results sent from BigQuery. Do not write PII or unprotected primary account numbers (PANs) 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.
VPC Service Controls is the Layer 7 perimeter service that prevents unauthorized copying or transferring of data from Google-managed services. A service perimeter is configured around in-scope projects containing sensitive data, and the perimeter then governs VM-to-service paths, service-to-service paths, and VPC ingress and egress traffic. Data flows from resources inside the perimeter to the outside only when an egress rule explicitly allows that destination, and requests flow from the outside to the inside only when an ingress rule allows that identity and client.
This service perimeter complements network isolation at lower layers. Virtual Private Cloud (VPC) networks and subnets provide Layer 3 and Layer 4 isolation for Compute Engine virtual machines (VMs) and services such as Google Kubernetes Engine (GKE). Cloud Armor filters IP allow and deny lists at the HTTP(S) load balancer at the edge of the Google network.
Exam tip: VPC networks isolate network traffic between VMs and subnets, while VPC Service Controls isolates data movement across Google-managed services with ingress rules for who can enter and egress rules for where data can leave.
Identity and Access Management (IAM) is the access-control system that decides which users, groups, and service accounts can act on sensitive data. A Google Cloud project is a trust boundary and IAM enforcement point, while folders group similar projects and apply folder-level IAM permissions that inherit down to their projects. Access follows the principle of least privilege, so designated users receive only the folders and projects they require.
Granular controls refine that access with IAM Conditions, principal access boundary (PAB) policies, and Privileged Access Manager for just-in-time temporary elevation. All users who require access to sensitive data must use multifactor authentication (MFA), implemented as 2-Step Verification, which supports PCI DSS requirements 7.2, 7.3, and 8.4.1 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 or sensitive authentication data. In Google Cloud the CDE also includes systems that provide security services such as IAM, systems that facilitate segmentation such as projects, folders, firewalls, VPCs, and subnets, and application pods and clusters that handle cardholder data. A system component counts as out-of-scope only when it is properly isolated from the CDE so that even if the out-of-scope component were compromised, the compromise would not impact the security of the CDE.
Segmentation on Google Cloud 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, with Organization policy constraints guarding against privilege escalation |
| Network segmentation | VPC networks and subnets | Place CDE resources in dedicated VPCs and subnets with firewall rules that deny all traffic by default and explicitly allow only required traffic |
| Service-level segmentation | Google-managed APIs and data | Define a VPC Service Controls perimeter around in-scope projects with ingress control for authorized identities and egress control for authorized destinations |
Without that isolation, the entire cloud footprint falls in scope, which increases assessment cost, control complexity, and attack surface.
Continuous compliance means showing 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, monitoring, and testing, but only when teams export, protect, and regularly review them with appropriate access controls on the destinations.
Comprehensive coverage for a PCI boundary combines network, perimeter, and access evidence:
Export audit logs to Cloud Storage for long-term archival and to BigQuery for analysis, monitoring, and forensics, and use Cloud Monitoring alert policies to notify responders when a security incident occurs.
Encryption strategy and key management for compliance is the plan for protecting sensitive commercial data while proving control over encryption keys to auditors. 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. Google Cloud encrypts customer data stored at rest and data moving in transit by default, with no action required, while regulated workloads use customer-managed keys, hardware protection, and external key storage for stronger proof of control.
Default encryption at rest means Google Cloud automatically encrypts all content stored at rest using Google-owned and Google-managed keys. The storage infrastructure encrypts user data before it is written to physical storage, 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 Google Cloud 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. This separation lets the service handle large volumes of data encryption while the customer controls the smaller wrapping key that ultimately grants or denies access.
A left-to-right process flow showing plaintext encrypted with a DEK, the DEK wrapped by a customer-managed KEK in Cloud KMS, and both stored together so disabling the KEK makes data unreadable.
Customer-managed encryption keys (CMEKs) are keys created and managed in Cloud KMS and used by integrated Google Cloud services to protect data at rest. Cloud KMS lets teams generate software keys, import existing keys, control rotation schedules, and use keys for encryption, decryption, signing, and client-side encryption. Cloud KMS Autokey automates provisioning and assignment by generating key rings, keys, and service accounts on demand during resource creation, instead of requiring them ahead of time.
The security team can enforce CMEK use with CMEK organization policies that block creation of non-compliant resources, while application developers remain responsible for selecting the mandated key on each resource. CMEK integrations encrypt data at rest across many services that commonly hold cardholder data and PII:
When a CMEK is disabled or destroyed, the service that depends on that key can no longer unwrap its DEKs, so the protected data becomes unreadable until access is restored.
Cloud HSM provides hardware-backed protection for Cloud KMS keys inside a Federal Information Processing Standard (FIPS) 140-2 Level 3 certified hardware security module (HSM). CMEKs can be software-based or HSM-protected, and HSM protection meets mandates that require key generation inside certified hardware and regular rotation. Cloud KMS also supports Single-tenant Cloud HSM instances for dedicated hardware isolation and Multi-tenant Cloud HSM keys for uses such as client-side encryption in Google Workspace.
Cloud External Key Manager (Cloud EKM) (opens in a new tab) lets teams store cryptographic key material outside Google Cloud in a compatible third-party key management system and link those external keys to Cloud KMS. Data still flows from the Google Cloud service to Cloud KMS for wrap and unwrap operations, but the key material itself remains under external custody and the service depends on that external system being available. Customer-supplied encryption keys (CSEK) are not recommended because Cloud EKM supports more services and higher availability for the scenarios CSEK historically addressed.
Exam tip: Accept default encryption when there is no mandate to manage keys, choose CMEK with Cloud KMS when compliance requires key lifecycle control, and choose Cloud EKM instead of CSEK when keys must stay outside Google Cloud.
Encryption in transit protects authenticity, integrity, and privacy when data moves outside physical boundaries controlled by Google. All virtual machine (VM)-to-VM traffic within a Virtual Private Cloud (VPC) network and between peered VPC networks is encrypted. Traffic from the internet or VPC network to Google services terminates at the Google Front End (GFE), which provides distributed denial-of-service countermeasures, routing, and load balancing.
Google infrastructure then uses Application Layer Transport Security (ALTS) for authentication, integrity, and encryption from the GFE to a Google Cloud service and from one Google Cloud service to another. Additional in-transit controls apply when defaults are not sufficient. Use MACsec for encryption over Cloud Interconnect connections and IPsec for encryption over Cloud VPN connections to on-premises data centers. Protect application-to-application traffic with Transport Layer Security (TLS) and mutual TLS through managed certificates on load balancers, Apigee, and Cloud Service Mesh for containerized applications. Compliance mappings for payment card data require strong cryptography in transit, including only trusted and valid certificates and minimum TLS 1.2 without fallback to insecure versions. Data in use in memory is not encrypted by default, so workloads handling confidential data can use Confidential Computing, which provides a trusted execution environment for Compute Engine instances through Confidential VMs.
An at-rest strategy for compliance balances auditability and control against cost and operational overhead. Default encryption is managed entirely by Google, CMEK shifts key lifecycle responsibility to the customer with central policy enforcement, and application-layer tokenization shifts responsibility to developers or a platform team and cannot be centrally enforced in the landing zone. Tokenization with Sensitive Data Protection replaces raw PII with tokens before storage in services such as BigQuery, so analysts can work without seeing sensitive fields, but the extra encrypt and decrypt pipeline adds cost, latency, and complexity.
| Strategy | Key ownership and control | Choose when |
|---|---|---|
| Default encryption at rest | Google-owned and Google-managed 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, HSM, and external-storage options, enforced by organization policy | Auditors require key management proof, FIPS 140-2 Level 3 hardware, or keys outside Google Cloud |
| Application-layer tokenization before storage | Application or pipeline-managed tokens plus underlying platform encryption | Highly sensitive fields must not be stored or used in raw form |
This choice directly supports commercial mandates such as Payment Card Industry Data Security Standard (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.
Commercial data protection on Google Cloud connects three controls into one flow. Sensitive Data Protection finds and de-identifies PII and cardholder data, VPC Service Controls plus IAM keeps the remaining sensitive data from leaving its projects and accounts, and Cloud KMS with TLS controls who can decrypt it at rest and who can read it in transit. Audit logs and data profiles then provide the shared evidence that the discovery, boundary, and encryption controls continue to work over time.
Virtual Private Cloud networks provide Layer 3 and Layer 4 isolation for Compute Engine virtual machines (VMs) and services such as Google Kubernetes Engine (GKE), while VPC Service Controls is a Layer 7 perimeter service that prevents unauthorized copying or transferring of data from Google-managed services. A VPC Service Controls perimeter governs data movement with ingress rules for who can enter and egress rules for where data can leave.
Use continuous discovery for profiling across an organization, folder, or project, and use on-demand inspection when you need the exact locations of personally identifiable information (PII) inside a specific BigQuery table or Cloud Storage bucket. Continuous discovery runs on its own without manual scans and produces data profiles with sensitivity level, data risk level, infoType findings, and location.
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. Masking keeps record structure while hiding the value with redacted symbols, tokenization keeps format and joinability without exposing raw values, and bucketing keeps statistical patterns by replacing precise values with broader ranges or categories.
The protected data becomes unreadable until access is restored because the service can no longer unwrap its data encryption keys (DEKs). In the envelope flow the service encrypts data with a DEK and Cloud KMS wraps that DEK with a customer-managed key encryption key (KEK), so access to data depends on access to both the encrypted data and the KEK.
An enterprise is designing an automated data ingestion pipeline to process transaction records before storing them in BigQuery for analytics. The incoming records include sensitive commercial and personally identifiable information (PII) with the following specific requirements:
Which combination of Sensitive Data Protection (Cloud DLP) de-identification transformations should the cloud architect specify in the de-identification template?
Use Format-Preserving Encryption (FPE-FFX) for credit card numbers, fixed-size bucketing for customer age, and an infoType inspection transformation with character masking for support notes.
Use BigQuery Dynamic Data Masking policy tags for credit card numbers, Cloud KMS envelope encryption for customer age, and Data Catalog column tagging for support notes.
Use Deterministic Encryption (DE) with AES-SIV for credit card numbers, date shifting for customer age, and cryptographic hashing for support notes.
Use Cryptographic Hashing (HMAC-SHA-256) for credit card numbers, primitive character masking for customer age, and replacement transformations for support notes.
Last updated