Study guide
9.1 IAM Policy Design: Verbs, Conditions, Dynamic Groups & Principals
OCI 2026 Architect Professional 1Z0-997-26
Track your progress as you read
Finished sections get a check mark.
1 Chapter 1: Exam Overview & Architecture Foundations
1.1 Exam Logistics, the 2026 Blueprint & How to Use This Guide 1.2 The OCI Well-Architected Framework & Architectural Tradeoffs
2 Chapter 2: High Availability & Resilient Design
2.1 Availability Domains, Fault Domains, Instance Pools & Load Balancers 2.2 OCI DNS Traffic Steering, Geolocation & Health Checks 2.3 Block Volume Replication, Volume Groups & Backup Policies 2.4 Object Storage Replication, Versioning & Retention Rules (WORM)
3 Chapter 3: Full Stack Disaster Recovery
3.1 Full Stack Disaster Recovery: Protection Groups, Members & Permissions 3.2 DR Plans: Switchover, Failover, Prechecks & Plan Execution 3.3 DR Drills, Plan Customization & Operational Readiness
4 Chapter 4: Container Instances, OCIR & OCI DevOps
4.1 Containerization, Microservices & OCI Container Instances 4.2 OCI Container Registry: Repositories, Push & Pull, and Image Management 4.3 OCI DevOps: Code Repositories, Build Pipelines & Deployment Strategies
5 Chapter 5: OCI Kubernetes Engine (OKE)
5.1 OKE Cluster Types, Node Options & Cluster Creation 5.2 OKE Networking: VCN-Native Pod Networking vs. Flannel, Subnet Sizing & Load Balancers 5.3 OKE Workload Identity, Secrets & Persistent Storage
6 Chapter 6: Serverless Functions & API Gateway
6.1 OCI Functions: Concepts, Invocation Modes, Integrations & Image Security 6.2 API Gateway: Authentication, JWT Validation, Rate Limiting & Usage Plans
7 Chapter 7: Messaging & Event-Driven Design
7.1 OCI Streaming, OCI Queue & Messaging Patterns for Microservices 7.2 OCI Events, Event Rules & OCI Notifications Fan-Out
8 Chapter 8: Infrastructure as Code with Terraform & Resource Manager
8.1 Infrastructure as Code with Terraform and the OCI Provider 8.2 OCI Resource Manager: Stacks, Jobs, Drift Detection & Private Endpoints
9 Chapter 9: Identity & Multi-Tenancy Governance
9.1 IAM Policy Design: Verbs, Conditions, Dynamic Groups & Principals 9.2 Organizations Management, Governance Rules & Compartment Quotas
10 Chapter 10: Security Posture with Cloud Guard & Security Zones
10.1 Cloud Guard: Detector Recipes, Problems, Scores & Responders 10.2 Security Zones & Vulnerability Scanning
11 Chapter 11: Secure Access & Network Protection
11.1 Layered Network Security: Security Lists, NSGs & Private Access to Services 11.2 OCI Bastion: Managed SSH, Port Forwarding & SOCKS5 Sessions 11.3 Zero Trust Packet Routing (ZPR): Security Attributes & Intent-Based Policy 11.4 OCI Network Firewall: Policies, Inspection Routing, Tunnel Inspection & Private NAT
12 Chapter 12: Application & Data Protection: WAF, Certificates & Vault
12.1 OCI Web Application Firewall: Policies, Modules, Actions & Enforcement Points 12.2 OCI Certificates: Private CAs, TLS Certificates, Renewal & Integrations 12.3 OCI Vault Key Management: Vault Types, Protection Modes, Dedicated KMS & External KMS 12.4 Vault Secrets: Versions, Rotation States, Automatic Generation & Rotation
13 Chapter 13: Base Database Service & Autonomous AI Database
13.1 Base Database Service: DB Systems, RAC, Data Guard, Cloning & Maintenance 13.2 Autonomous AI Database: Workload Types, Deployment Options, Services & Connectivity 13.3 Operating Autonomous AI Database: Scaling, Cloning, Backups, Patching & Self-Tuning 13.4 Autonomous Data Guard & Backup-Based Disaster Recovery for Autonomous AI Database
14 Chapter 14: Recovery Service, Global Distribution & OCI Cache
14.1 Autonomous Recovery Service: Protection Policies, Real-Time Protection & Retention Lock 14.2 Globally Distributed Autonomous AI Database: Sharding, Data Placement & Replication 14.3 OCI Cache: Valkey and Redis Clusters, Sharding, Security & Cross-Region Replication
15 Chapter 15: Monitoring, Alarms & Notifications
15.1 OCI Monitoring: Metrics, Namespaces, Custom Metrics & Monitoring Query Language 15.2 Alarms & Notifications: Trigger Rules, Absence Alarms, Split Messages, Suppression & Delivery
16 Chapter 16: Logging, Audit, Connector Hub & Search
16.1 OCI Logging: Log Groups, Service Logs, Custom Logs & the Unified Monitoring Agent 16.2 Logging Search, the Logging Query Language & OCI Audit 16.3 Connector Hub: Moving Logs, Metrics, Queue Messages & Streams Between Services 16.4 OCI Search with OpenSearch & Resource Search for Observability
9.1 IAM Policy Design: Verbs, Conditions, Dynamic Groups & Principals
Key Takeaways Explain This Topic AI
OCI IAM policy statements use the form Allow subject to verb resource-type in location where conditions, and access is denied unless a policy allows it.
The OCI verb hierarchy defines four progressive levels of capability: inspect (metadata listing only), read (inspect plus resource content viewing), use (read plus resource modification without creation/deletion), and manage (full administrative lifecycle control).
Where-clause conditions provide granular security controls, enforcing source IP network perimeters, mandatory multi-factor authentication (request.user.mfaTotpVerified = 'true'), and Attribute-Based Access Control (ABAC) using defined tags.
Deny policies are an opt-in feature that administrators must enable, and members of the Administrators group in the default domain are always exempt from them.
Last updated: October 2026
- 7.1 IAM Policy Design: Verbs, Conditions, Dynamic Groups & Principals
- The Anatomy of Advanced IAM Policy Syntax
Oracle Cloud Infrastructure uses a declarative policy language built on least privilege. All access is denied by default , and a principal can do only what allow statements grant. Allow statements are additive across all applicable policies in the tenancy and compartment hierarchy.
OCI also offers deny policies as an opt-in feature that an administrator must enable. Deny statements use the same syntax but begin with deny (also deny admit and deny endorse for cross-tenancy statements). They explicitly block actions, for example to stop anyone except a security team from changing certain policies. Members of the default domain's Administrators group are always exempt from deny policies, so the tenancy can never lock itself out.
Every policy statement adheres to a formal syntactic structure: Allow <subject> to <verb> <resource-type> in <location> where <conditions>
Understanding each component of this grammar is essential for enterprise governance:
<subject> : Identifies the principal being authorized. This can be a domain-qualified group ( group IdentityDomainName/GroupName or group GroupName within the evaluating domain), a dynamic group ( dynamic-group DynamicGroupName ), or all authenticated tenancy users ( any-user ).
<verb> : Specifies the level of access granted ( inspect , read , use , manage ). Verbs operate cumulatively.
<resource-type> : Identifies the cloud resource or family of resources being targeted (e.g., virtual-network-family , instance-family , buckets , secret-bundles ).
<location> : Defines the compartment boundary where permissions apply. This is expressed as tenancy (granting access across all compartments in all subscribed regions) or compartment <compartment-name> . For nested compartment hierarchies, paths are expressed using colon notation: compartment ParentCompartment:ChildCompartment or by targeting a specific OCID: compartment id <compartment-ocid> .
where <conditions> : An optional clause that refines authorization using boolean expressions and environmental context.
Verb Hierarchy and Granular Resource Types
OCI structures access through an ordered verb hierarchy . Each ascending verb level inherits all capabilities of the verbs below it:
inspect : The lowest access tier. Grants permission to discover and list resources of the specified type without viewing user data, sensitive attributes, or resource contents. For example, inspect instances allows listing compute instance OCIDs, display names, and lifecycle states, but prohibits viewing instance metadata, VNIC IP attachments, or console history.
read : Encompasses inspect and adds permission to view user-defined metadata and retrieve actual resource contents. For instance, read buckets allows listing and downloading objects from an Object Storage bucket, while read instances exposes attached volume details and IP configurations. The read verb does not permit modifying, creating, or terminating resources.
use : Encompasses read and adds permission to work with existing resources, modify operational parameters, and execute operational state changes. For example, use instances allows starting, stopping, and rebooting existing virtual machines, and attaching block volumes. However, use prohibits creating new resources or permanently deleting/terminating existing resources .
manage : The highest tier. Grants complete administrative control, encompassing inspect , read , use , plus the creation and permanent destruction (deletion/termination) of resources.
Individual vs. Family Resource Types
OCI organizes resource types into individual types (fine-grained control over a specific resource) and family types (aggregate groupings encompassing all related subordinate types):
Individual resource examples: vcns , subnets , route-tables , security-lists , instances , instance-images , buckets , objects .
Family resource examples: virtual-network-family , instance-family , object-family , database-family , all-resources .
Enterprise security architecture mandates granting permissions using individual resource types whenever possible. Granting manage virtual-network-family to application developers allows them to alter core routing tables and Internet Gateways, whereas granting use subnets allows them to attach virtual network interfaces (VNICs) to existing subnets without granting the power to alter network security architecture.
OCI IAM Policy Verb Hierarchy & Operational Scope
Verb Level Inherited Verbs Granted Capabilities Prohibited Operations Enterprise Architectural Use Case
inspect None List resources, view OCIDs, check operational lifecycle status Cannot view user metadata, object contents, or configurations Third-party cloud cost auditing tools, inventory scanners
read inspect View user metadata, download object contents, inspect configurations Cannot alter state, update parameters, create, or delete Security auditors, compliance monitors, read-only analysts
use inspect , read Work with existing resources: start/stop instances, attach volumes/VNICs, most updates Cannot create or delete; updates that act like a create (such as UpdateSecurityList or UpdatePolicy) need manage Daily application operators, support engineers, patch admins
manage inspect , read , use Full lifecycle governance: create, update, reconfigure, delete None (full administrative entitlement within compartment) Cloud platform engineering leads, DevOps automation pipelines
Where-Clause Conditions: Network Scoping, MFA Enforcement & ABAC Tags
Where-clause conditions allow architects to inject contextual zero-trust evaluations into policy decisions. Conditions evaluate request context ( request.* ) and target resource attributes ( target.* ).
1. Network Perimeter Enforcement
Policies can restrict API calls to requests that come from approved networks. You first create a network source (a named set of public CIDR blocks and/or VCN addresses), then reference it in a condition: Allow group SecurityAdmins to manage all-resources in tenancy
- where request.networkSource.name = 'corpnet'
- 2. Mandatory Multi-Factor Authentication (MFA)
To protect mission-critical environments against compromised API signing keys or stolen user sessions, architects can mandate that the caller has completed multi-factor authentication: Allow group ProductionOps to manage database-family in compartment Production
where request.user.mfaTotpVerified = 'true'
If a user authenticates without completing an MFA challenge (such as invoking an API script with an un-stepped session), OCI denies the request with an authorization error (often reported as NotAuthorizedOrNotFound).
3. Attribute-Based Access Control (ABAC) Using Defined Tags
Role-Based Access Control (RBAC) often leads to policy explosion in large enterprises. OCI solves this through Attribute-Based Access Control (ABAC) , using Defined Tag namespaces to govern access dynamically: Allow group Developers to manage instance-family in compartment Development
where target.resource.tag.Project.Environment = 'NonProduction'
Furthermore, architects can enforce dynamic ownership matching between the calling user's group and the target resource: Allow group CostCenterAdmins to manage all-resources in compartment SharedWorkloads
where request.principal.group.tag.Finance.CostCenter = target.resource.tag.Finance.CostCenter
This single policy dynamically authorizes hundreds of project groups across the tenancy without requiring individual group-to-compartment policy statements.
Dynamic Groups and Resource Principals Architecture
Traditional architectures frequently embed static administrative credentials, OCI API signing keys, or private SSH keys inside compute virtual machines, container configurations, or application source code. If code repositories leak or an instance is compromised, static credentials provide adversaries with uninhibited lateral movement across the tenancy.
OCI eliminates static machine credentials through Dynamic Groups and Resource Principals .
Dynamic Groups
A Dynamic Group is an IAM principal whose membership is determined dynamically at runtime based on matching rules evaluated against resource metadata. Instead of adding human users, architects define membership criteria for OCI resources: All {instance.compartment.id = 'ocid1.compartment.oc1..aaaaaaaaxxx', tag.Operations.Tier = 'Backend'}
Membership rules support logical operators: Any { ... } (OR) and All { ... } (AND). Any compute instance, autonomous database, or function meeting the rule automatically inherits all IAM policies granted to that dynamic group.
Resource Principals
Resource Principals extend this paradigm to serverless, containerized, and managed services. When an OCI Function, OKE Pod (via Workload Identity), OCI Data Flow application, or OCI Data Science job invokes an OCI API, it does not supply an API signing key. Instead, the application utilizes the OCI SDK configured for Resource Principal authentication:
OCI uses several principal types for non-human callers:
Instance principals : a compute instance in a dynamic group obtains short-lived certificates from the instance metadata service and signs API calls with them. Every process on the instance shares that identity.
Resource principals : services such as OCI Functions, Data Flow, Data Science jobs, Full Stack DR protection groups, and DevOps pipelines receive a resource principal session token from the service environment (for Functions, the runtime injects it into the container). Matching rules use resource.type , for example resource.type = 'fnfunc' .
OKE workload identity : pods in enhanced clusters authenticate as a specific cluster, namespace, and service account.
In every case the SDK obtains and refreshes short-lived credentials automatically, and IAM evaluates the policies written for the dynamic group or workload: Allow dynamic-group PaymentProcessingFunctions to read secret-bundles in compartment Finance
Because these credentials are short-lived and issued automatically, there are no API keys to store, rotate, or leak from source repositories.