Prepare and test your skills

Prepare and test your skills

Defense in depth in Google Cloud (opens in a new tab) is a security strategy that stacks multiple independent layers so failure of one control does not expose data. The layers span physical data centers and custom hardware, the global network, data encryption, identity and access, and security operations. Google secures the underlying cloud infrastructure while the customer secures data and how data is accessed and used.
A layered stack diagram showing Google's defense-in-depth model: operations and monitoring on top, then identity and access, data encryption, the global network, and hardware/physical security at the foundation, all protecting a core of customer data. Each independent layer means a failure in one control does not expose the data.
The hardware and physical layer protects data centers, servers, storage, and custom chips that run Google Cloud. Perimeter controls include fencing, vehicle barriers, metal detectors, alarms, electronic access cards, laser intrusion detection, and 24/7 cameras, plus guards who pass background checks. Entry to the data center floor requires a security corridor with badge and biometric checks open only to approved roles. Inside, Google reduces risk with custom servers that omit unneeded video cards and peripheral connectors, vetted component vendors, and Titan security chips (opens in a new tab) that verify boot code and store keys. Redundant power, backup generators, cooling, and fire protection support continuous operation, while retired drives are sanitized and verified or shredded and recycled securely.
The network layer controls how traffic moves between users and Google Cloud resources. Google's global network connects directly to most internet providers, which limits hops across the public internet and reduces interception chances. Only authorized services and protocols traverse the network and all else is dropped, while traffic flows from customers through Google Front End (GFE) servers to backends where GFE stops malicious requests and distributed denial-of-service (DDoS) attacks. The data layer encrypts customer data at rest by default and traffic in transit with Transport Layer Security (TLS), while Confidential Computing protects data in use inside isolated hardware. The identity layer follows zero trust, so every request is authenticated and authorized with multi-factor authentication (MFA), least privilege, and checks of role, device, and location.
Exam tip: Google Cloud encrypts data at rest and in transit by default, but data in use is not encrypted by default; Confidential Computing protects data in use.
Shared responsibility defines a clear boundary where Google protects the cloud itself and the customer protects what the customer brings to the cloud. Because encryption, network defense, and monitoring are built in and on by default, organizations buy fewer add-on tools and spend less on physical security and low-level patching.
| Cloud model | Google secures | Customer secures |
|---|---|---|
| Infrastructure as a service (IaaS) | Hardware, storage, and network | Data, applications, operating systems, and access |
| Software as a service (SaaS) | Everything except data and its access and usage | Data and how data is accessed and used |
Redundancy of servers, storage, networks, and geographically distributed data centers supports continuity during regional disruptions. Independent verification against ISO/IEC 27001, SOC 2, SOC 3, and FedRAMP helps regulated industries meet compliance needs.
Vertically integrated infrastructure in Google Cloud (opens in a new tab) means Google designs its own data centers, servers, network equipment, and security chips so performance, efficiency, and protection are built together. This control improves security, lowers latency, increases availability, and reduces energy use compared with assembling unrelated parts.
The global private network is Google's worldwide IP network built from Google-owned fiber, public fiber, and undersea cables. Because the network connects directly to most internet service providers in the world, customer traffic enters Google's protected network early instead of traveling long distances over the public internet. Data normally travels in hops between providers, and each avoided hop removes another chance for interception. Security on this network uses defense in depth. Only authorized services and protocols traverse the network, firewalls and access control lists enforce segregation, and traffic flows from customer devices to Google Front End (GFE) servers at the edge where harmful requests and DDoS attacks are filtered before reaching production systems.
Exam tip: Fewer public internet hops means less exposure — Google limits hops by linking its private network directly to most ISPs.
Google data centers are custom facilities designed for power usage effectiveness (PUE), a measure of how efficiently electricity is used. Every critical power component has primary and alternate sources, backup generators can run a data center at full capacity, and cooling and fire detection maintain safe operating temperature. Google was the first major internet company to gain external certification for environmental, safety, and energy management across data centers, including voluntary ISO 50001 in Europe. Custom servers omit unneeded video cards, chipsets, and peripheral connectors that waste power and can introduce vulnerabilities. Component vendors are vetted and audited, resources are dynamically allocated to meet demand, and software monitors for binary changes with automated self-healing. Storage devices use full disk encryption (FDE) and drive locking, and retired drives are verified as sanitized or physically shredded and recycled.
A hardware root of trust is a protected hardware starting point that is trusted before software starts. The Titan security chip (opens in a new tab) provides this anchor and verifies boot firmware in order, so each verified stage passes trust to the next stage and altered code cannot start silently. Using the same chip globally gives every server a consistent, known-good starting state. Migration to this efficient platform reduces energy use and emissions by 1.4 to 2 times compared with typical on-premises deployments:
| Environment | Energy and emissions | Cooling and power design |
|---|---|---|
| Typical on-premises data center | Higher energy use and emissions | Often lacks scale for advanced cooling and power systems |
| Google Cloud data center | 1.4 to 2 times lower when workloads migrate | Modern facility built for PUE |
Google is responsible for sustainability of the cloud, meaning efficient data centers and infrastructure, while customers are responsible for sustainability in the cloud by right-sizing resources and using services that scale to zero.
Encryption in Google Cloud (opens in a new tab) protects data in all three states — at rest, in transit, and in use — by making content unreadable without the correct keys. Data at rest is encrypted by default, data in transit is encrypted as it moves between users and services, and data in use requires special isolation while it is processed in memory. Key management decides who controls decryption and where keys live.
Default encryption protects customer data at rest in production data centers with no customer action required. Encryption is applied at the application layer, the storage device layer, or both, using keys held in Keystore, Google's internal key service, that are rotated automatically by most services. Customers cannot view Keystore keys or their usage logs, and keys may be shared across customers, so default encryption prevents accidental unencrypted storage but gives no control over location or destruction. Organizations that need control can choose stronger options:
| Option | Who manages keys | Where key material lives | Customer control |
|---|---|---|---|
| Google default encryption | Google in Keystore | Google infrastructure | Automatic rotation; no key visibility |
| Customer-managed encryption keys (CMEKs) | Customer in Cloud KMS | Cloud KMS as software or HSM-protected keys | Location, access, rotation, use, and destruction |
| Customer-supplied encryption keys (CSEKs) | Customer provides material per request | In memory; not permanently stored by Google | Direct control of key material; Compute Engine support deprecated |
| Cloud External Key Manager (Cloud EKM) | Customer in third-party system | Outside Google infrastructure | Separation of data and keys while using Google services |
Cloud KMS Autokey can automatically provision CMEKs for new resources when its defaults meet requirements.
Encryption in transit combines authentication that verifies each peer, integrity that detects alteration, and confidentiality that keeps data unreadable during travel. Together they remove the need to trust the lower network layers, so intercepted traffic cannot be read or quietly changed. Traffic flows from users on the internet to Google Front End (GFE), which terminates HTTP(S), TCP, and Transport Layer Security (TLS) connections at points of presence worldwide and routes traffic over Google's network to the destination. GFE negotiates forward-secure TLS with modern clients and the strongest legacy option with older clients, using BoringSSL validated to FIPS 140-3 Level 1. Inside Google, VM-to-VM traffic within a Virtual Private Cloud (VPC) network or across peered VPCs is encrypted at the network layer, while Application Layer Transport Security (ALTS) provides mutual authentication and AES-128-GCM encryption between GFE and services and between services.
Exam tip: Connections between VMs that use internal IP addresses are automatically encrypted, while connections that use external IP addresses leave Google's network and are not automatically encrypted.
Confidential Computing is the Google Cloud family that protects data in use by running computation inside a Trusted Execution Environment (TEE) with hardware isolation and attestation. The TEE keeps the workload cryptographically isolated in multi-tenant hardware, prevents unauthorized access during processing, and produces verifiable proof of system state. Options include Confidential VM, Confidential Google Kubernetes Engine Nodes, Confidential Dataflow, Confidential Managed Service for Apache Spark, and Confidential Space for joint analysis. Cloud Key Management Service (Cloud KMS) creates, stores, rotates, and controls keys with envelope encryption, with software keys, Cloud HSM hardware-backed keys, and Cloud EKM external keys. Key Access Justifications adds a justification record every time an external key is used and requires customer approval before Google personnel can decrypt data at rest.
Authentication, authorization, and auditing form a connected chain that proves identity, grants permitted actions, and records what happened. Authentication (opens in a new tab) verifies who a user or workload is, authorization decides what that verified identity can do, and auditing records who did what, where, and when. Identity management creates the trusted identities that make the other two steps possible.
Identity management covers provisioning, managing, and deprovisioning corporate identities for employees, customer identities for website users, and service identities for applications. Corporate identities are centrally managed with Cloud Identity and Google Workspace so sign-in rules are consistent and users move away from uncontrolled personal accounts. A principal must authenticate with a recognized identity before any authorization check can succeed.
People sign in with Cloud Identity or Google Workspace credentials for console, Cloud Shell, and gcloud CLI work after running gcloud auth login. Trust is strengthened with strong passwords, multi-factor authentication (MFA) for all users, and federation that lets users sign in with an external provider such as Active Directory. Workloads prove identity with Google Cloud service accounts attached as default identities, or with Workload Identity Federation that lets code on Google Cloud, AWS, Azure, or GitHub use external identities with short-lived tokens.
| Federation option | How sign-in works | Accounts needed |
|---|---|---|
| Synced federation with Cloud Identity | External identities sync to matching Cloud Identity accounts | External account plus Cloud Identity account |
| Workforce Identity Federation | External provider signs users directly into Google Cloud | External account only |
Authorization and access control (opens in a new tab) in Google Cloud is enforced by Identity and Access Management (IAM), which checks permissions on every action. A permission such as storage.objects.create allows one operation, while a role collects permissions and is granted to a principal on a resource. Permissions cannot be granted directly; a principal receives access only through a role binding in an allow policy.
The resource hierarchy flows from organization to folders to projects to service resources, and a child inherits allow policies from its parents. An IAM deny policy overrides any allow policy on the same resource and descendants, except deny policies cannot be set on Cloud Storage buckets or objects.
| Role type | Who manages permissions | Choose when |
|---|---|---|
| Predefined role | Google Cloud maintains set | Work matches a common task |
| Custom role | Customer selects permissions | Predefined roles grant too much or too little |
| Basic role | Google provides broad Owner, Editor, Viewer | Short-term testing in non-production only |
Least privilege means granting the smallest role at the lowest resource level, using separate service accounts per component, and using conditional bindings and temporary access so sensitive rights expire.
Exam tip: A deny policy always overrides an allow policy, so an explicit denial blocks access even when a role grant would otherwise allow it.
Cloud Audit Logs creates an immutable record that cannot be changed after writing, which makes logs trustworthy for forensics and compliance. Four types cover each project, folder, and organization:
Admin Activity and Policy Denied logs are especially important for access work because they show what changed and what was blocked. Logs are viewed in Logs Explorer or the Cloud Logging API, retained for a fixed period, and exported with log sinks to BigQuery for long-term analysis. The Logs Viewer role reads Admin Activity, Policy Denied, and System Event logs, while Private Logs Viewer also reads Data Access logs in the _Default bucket.
Continue with this study guide: Differentiate between authentication, authorization, and auditing.
Identity and Access Management (IAM) and Two-Step Verification (2SV) reduce unauthorized access by proving identity strongly and granting only needed permissions. Cloud Identity centrally manages users and groups and proves who a user is, while Identity and Access Management (IAM) controls what that identity can do on which resource. Together they limit attack surface and improve productivity.
A principal is any identity that can request access, and no principal can act until IAM permits it. IAM recognizes several principal types:
A role is a collection of permissions where each permission allows one operation, such as creating a virtual machine. Job-function roles group permissions for common jobs so administrators assign one role instead of many permissions. An IAM allow policy binds principals to a role on a specific resource, so who gets access, what is allowed, and which asset is affected are defined in one place. The Organization resource sits at the apex of the hierarchy and provides central visibility over folders, projects, and service resources such as virtual machines, buckets, and topics.
Least privilege means every principal receives only minimum permissions at the narrowest scope needed for the task. A role cannot grant a single permission directly, so administrators choose the smallest predefined or custom role and attach it at the lowest level such as an instance or repository rather than the whole project. If that account is compromised, it can reach fewer resources. Secure Source Manager shows this pattern with separate instance roles and repository roles, so access to an instance does not automatically grant access to every repository. Cloud Infrastructure Entitlement Management (CIEM) finds federated identities from Entra ID, Okta, or Active Directory with excess roles and surfaces remediations in Security Command Center. OS Login enforces least privilege for SSH by checking IAM permissions on every connection and tying keys to account lifecycle, unlike metadata-based SSH keys checked only at upload.
Exam tip: OS Login checks IAM permissions every time a user starts an SSH session, while metadata-based SSH keys are checked only when the key is uploaded.
Two-Step Verification (2SV) requires two separate proofs before granting access: something the user knows, typically a password, plus something the user has, such as a phone or physical key. Passwords alone can be phished, guessed, or stolen by malware, and a stolen password lets an attacker sign in from anywhere. The second factor breaks this pattern because the attacker does not possess the extra device. Sign-in follows a strict order: first the password is verified, then the second factor is verified, and only then access is granted. The second step depends on the first succeeding, and access depends on both succeeding.
| Second factor | What user needs | How verification works |
|---|---|---|
| Physical security key | Hardware key registered to account | User taps key, which cryptographically confirms real site |
| Mobile prompt | Trusted smartphone signed into account | User taps Yes to approve sign-in attempt |
Physical keys resist phishing because they will not respond to a fake site, while mobile prompts block most credential theft with easy approval. Emergency access accounts remain as a tightly limited exception, with two to five super admins protected by FIDO keys and audited activity.
Continue with this study guide: Describe the benefits of using two-step verification (2SV) and IAM.
Network attack protection in Google Cloud filters unwanted traffic at the edge before it reaches backends. A client request first arrives at Google's edge, the Google Cloud Armor (opens in a new tab) security policy checks it against allow and deny rules, and only permitted traffic flows to the backend. This edge-first order saves backend capacity and keeps services available during large-scale attacks.
A left-to-right process flow showing client traffic arriving at Google Front End servers, where a Cloud Armor security policy checks IP, geography, and WAF rules; allowed traffic passes through the Application Load Balancer and VPC firewall to backends, while denied traffic is dropped at the edge.
A Google Cloud Armor security policy is a set of rules that allows or denies incoming traffic based on request characteristics. The policy acts at Google Front Ends (GFEs), Google's global entry points, so malicious requests never contact backends. Filters can be used alone or together:
Policies attach to load balancer backend services behind an external Application Load Balancer. Traffic flows from the client to GFE, through the policy check, then from the load balancer to the backend only if allowed. The same policy can be reused across multiple backend services in complex URL maps, or different policies can protect different services.
A denial-of-service (DoS) attack from one source and a distributed denial-of-service (DDoS) attack from many compromised hosts both try to make a service unavailable by overwhelming it. Externally exposed APIs are especially vulnerable because they must accept internet traffic. Defense must therefore happen before traffic reaches the application. The global edge terminates user traffic at GFE and automatically scales to absorb attacks such as SYN floods before they reach Compute Engine instances. Cloud Load Balancing uses a single anycast IP to front instances in all regions and directs each user to the closest backend with capacity, so during an attack load shifts to regions with available capacity. On top of this capacity, Cloud Armor adds Web Application Firewall (WAF) rules from the ModSecurity Core Rule Set aligned with OWASP Core Rule Set 4.22 to block cross-site scripting (XSS), SQL injection (SQLi), and beta rules for remote file inclusion (RFI), local file inclusion (LFI), and remote code execution (RCE).
Exam tip: Cloud Armor security policies act at the edge on client traffic to GFEs, while VPC firewall rules control traffic to and from backends.
Cloud Armor, Cloud CDN, and Virtual Private Cloud (VPC) firewalls act at different points in a fixed order. Google infrastructure absorbs volume at the edge, Cloud Armor filters edge traffic, and backend firewalls enforce the final boundary.
| Control | Where it acts | What it controls |
|---|---|---|
| Google Cloud Armor security policy | Edge on client traffic to GFEs | Which external clients pass, by IP, geography, and WAF rule |
| VPC firewall rules | Backend network to and from VMs | Which load balancer and health-check traffic reaches backends |
Traffic moves through both controls in order:
When Cloud Armor works with Cloud CDN, policies enforce only for dynamic content, cache misses, or origin requests and do not apply to cache hits. Metrics appear in Cloud Monitoring and suspicious patterns appear in Security Command Center.
Cloud SecOps in Google Security Operations (opens in a new tab) integrates continuous monitoring, threat detection, investigation, and response into one cloud-native workflow. Google SecOps is a private customer layer on Google infrastructure where the enterprise privately retains, analyzes, and searches months of telemetry while Google manages scale and protection. Credentials for feeds are stored in Secret Manager and access is controlled with Identity and Access Management (IAM).
A threat moves through Google SecOps in a fixed lifecycle where each stage feeds the next:
Collection uses a forwarder for syslog and packet capture, a collector for many sources, ingestion APIs including the Chronicle API for direct sends, and data feeds for Amazon S3, Okta, Microsoft 365, and Microsoft Entra ID plus direct pulls of Cloud Logging, Cloud Asset Inventory, and Security Command Center findings. Detection parses raw logs into UDM while keeping the original raw log, indexes UDM for fast search, enriches it with entity, geolocation, and WHOIS context, and runs the Detection Engine with Google-built and custom rules.
Exam tip: Google SecOps keeps both versions of telemetry: the normalized UDM event for consistent searching and the original raw log for verification.
Investigation answers what a threat is, what it is doing, whether it matters, and how to respond. Analysts search across all domains and narrow to a specific asset, domain, or IP, using UDM Search for normalized events, Raw Log Scan for unparsed logs, and regular expressions for pattern matching. Asset, user, IP, domain, and hash views show impact on the enterprise, while graph investigator visualizes relationships and prevalence graphs show how many domains an asset contacted over time. Response combines human collaboration with automation. Case management groups related alerts, supports triage and assignment, and retains audit history, while dashboards track SOC metrics. Playbook Designer builds workflows by dragging predefined actions onto a canvas without coding, and an integrated development environment (IDE) lets engineers modify actions and build new integrations. Prebuilt integrations send remediation to endpoint, blocking, and ticketing systems.
The cloud-native foundation divides work while sharing the same data:
| Platform part | Focus | Role in Google SecOps |
|---|---|---|
| SIEM capabilities | Retain, analyze, and search telemetry | Normalize, index, correlate, and support long-term search |
| Security Orchestration, Automation, and Response (SOAR) capabilities | Orchestrate and automate workflows | Unify ingestion and run playbook actions as central execution environment |
Because both parts run on Google infrastructure, monitoring operates continuously without managing storage servers. Centralization reduces tool switching, automation shortens time from alert to resolution, and months of history preserve context for hunting. Compliance is strengthened with Assured Workloads for data residency, FedRAMP High and Moderate and DoD Impact Levels 2, 4, and 5, HIPAA and PCI DSS support, customer-managed encryption keys (CMEK), and VPC Service Controls perimeters against exfiltration.
Defense in depth in Google Cloud is a security strategy that stacks multiple independent layers so failure of one control does not expose data. The layers span physical data centers and custom hardware, the global network, data encryption, identity and access, and security operations. Google secures the underlying cloud infrastructure while the customer secures data and how data is accessed and used.
Use Customer-managed encryption keys (CMEKs) instead of Google default encryption when you need control over key location, access, rotation, use, and destruction. Google default encryption protects customer data at rest automatically with keys held in Keystore that are rotated automatically but cannot be viewed by customers and may be shared across customers. CMEKs are managed by the customer in Cloud Key Management Service (Cloud KMS) as software or hardware security module-protected keys.
Google Cloud Armor security policies filter external client traffic at the edge on traffic to Google Front End (GFE) servers by IP, geography, and Web Application Firewall rules, while Virtual Private Cloud (VPC) firewall rules control traffic to and from backend virtual machines. Allowed traffic flows from the client to GFE through the policy check and then through the load balancer to the backend, where ingress allow rules permit only that load balancer and health-check traffic to reach backend virtual machines.
Last updated