Cheat sheet
Microsoft SC-500 Cheat Sheet
Identity, Access, Governance
- 20-25% of exam
- PIM Essentials Conditional Access Key Vault Protection Azure Policy Effects RBAC and Locks Identity Control Picker
Storage, Databases, Networking
- 25-30% of exam
- Storage Account Security Azure SQL Security NSG Default Rules Firewall Rule Order Private Access Options Network Control Picker
Secure Compute and AI
- 20-25% of exam
- Security for AI AI Guardrails APIM AI Gateway VM and Server Hardening Containers and App Platform AI Control Picker
Manage and Monitor Posture
- 20-25% of exam
- CSPM Plan Split Defender Workload Plans Sentinel Setup Sentinel Data Collection Security Copilot Defender Plan Picker
Quick Facts
- Exam SC-500
- Credential Cloud and AI Security Engineer
- Level Associate
- Time 120 min
- Pass 700/1000
- Fee $165 USD
- Items Not published (40-60 typical)
- Delivery Pearson VUE, proctored
- Renewal 1 year, free online
- Replaces AZ-500
- Blueprint May 13 2026
Key Vault Retention ID
- Soft delete 7-90 days | default 90
- Soft delete cannot be disabled Purge protection is optional Set once at creation
Policy vs Lock ID
Azure Policy
- Allow, deny, remediate
- Audits compliance
- Scales by scope
Resource lock
- CanNotDelete or ReadOnly
- Stops accidental change
- Inherited by children
- Standard vs protection
Identity Control Picker ID
- Standing admin rights → PIM eligible role ( Just-in-time )
- Risky sign-in → Conditional Access ( Risk condition )
- App calls Azure → Managed identity ( No stored secret )
- Identity shared by apps → User-assigned identity ( Own lifecycle )
- App holds a secret → Key Vault ( Use RBAC model )
- Stop accidental deletion → Resource lock ( CanNotDelete )
- Enforce required tags → Azure Policy ( deny effect )
- Too many role assignments → Azure RBAC review ( Least privilege )
- Prove compliance state → Defender for Cloud ( Regulatory compliance )
PIM Essentials ID
- Eligible Must activate before use JIT
- Active Rights already granted
- Activate MFA, justification, approval
- Time-bound Start and end dates
- Permanent eligible Always able to activate
- Approver Approves or denies request
- Access review Confirms access still needed
- Audit history Downloadable activation record
- PIM scope Entra roles, Azure roles, groups
Conditional Access Formula ID
- Assignments + conditions -> grant or session
- Assignments: who and what Conditions: risk and device Grant: block or require Session: limit the experience
Eligible vs Active ID
Eligible
- Activation required
- MFA and justification
- Expires after activation
Active
- Rights already usable
- No activation step
- Can still be time-bound
- Request it vs have it
Conditional Access ID
- Assignments Who, what, where
- Users and groups Include or exclude
- Target resources Apps, actions, auth context
- Network IP ranges and geographies
- Conditions Risk, platform, client app
- Grant Block or require controls
- Session Sign-in frequency, app restrictions
- Report-only Evaluate without enforcing
- Many policies All must be satisfied
- Block control Wins over every grant
Key Vault RBAC vs Access Policy ID
Azure RBAC
- Recommended and default
- Granting needs Owner
- Works with PIM
Access policy
- Legacy data-plane only
- Vault-wide permissions
- Contributor can self-grant
- Central vs vault-local
App and Managed Identities ID
- App registration Tenant application object
- Enterprise app Service principal instance
- System-assigned Tied to resource lifecycle
- User-assigned Standalone, shareable identity
- OAuth grant Delegated permission consent
- Consent settings Limit what users approve
- Admin consent workflow Users request admin approval
Key Vault Protection ID
- Soft delete On by default, permanent
- Retention range 7 to 90 days
- Default retention 90 days
- Retention set once Fixed at vault creation
- Purge protection Optional, blocks early purge
- Key Vault Purge Operator Role allowed to purge
- RBAC model Recommended and now default
- Access policy Legacy data-plane model
- Vault firewall Deny public, allow trusted
- Defender for Key Vault Vault access threat alerts
Azure Policy Effects ID
- deny Blocks noncompliant creation
- audit Flags but allows
- auditIfNotExists Checks related resource
- deployIfNotExists Deploys missing configuration
- modify Changes properties or tags
- append Adds required fields
- denyAction Blocks a named action
- disabled Turns the rule off
- Order Append, modify, deny, audit
- Initiative Grouped policy definitions
RBAC and Locks ID
- Owner Full access plus granting
- Contributor Manages, cannot grant access
- User Access Administrator Assigns roles only
- Reader View only
- Custom role Actions plus notActions
- Deny assignment Blocks despite role grant
- CanNotDelete lock Read and edit allowed
- ReadOnly lock No edits or deletes
- Lock inheritance Child scopes inherit locks
Firewall Order DA
- Threat intel -> DNAT -> Network -> Application
- Threat intel first DNAT before network Application rules last Match stops processing
NSG vs ASG DA
NSG
- Allow or deny rules
- Subnet or NIC
- Priority 100-4096
ASG
- Named group of NICs
- Used inside NSG rules
- No rules of its own
- Rule vs grouping
Network Control Picker DA
- Filter subnet traffic → NSG ( Layer 3-4 )
- Group VMs by role → ASG ( Reusable source )
- Enforce rule tenant-wide → Security admin rule ( Beats NSG )
- Filter outbound FQDN → Azure Firewall ( Application rule )
- Inspect encrypted egress → Firewall Premium ( TLS inspection )
- Block SQL injection → WAF ( Prevention mode )
- Reach PaaS privately → Private endpoint ( Private IP )
- Offer your own service → Private Link service
- Retire the VPN → Entra Private Access ( Per-app access )
- Rule not applying → Network Watcher ( Effective rules )
Storage Account Security DA
- User delegation SAS Signed by Entra credentials Best
- Service SAS One service, account key
- Account SAS Many services, account key
- Stored access policy Revoke a service SAS
- Disable Shared Key Forces Entra authorization
- Storage firewall Default deny public networks
- Trusted services Exception for Azure platform
- Defender for Storage Malware and exfiltration alerts
- Sensitive data detection Prioritizes alerts by sensitivity
NSG Default Priorities DA
- 65000 VNet | 65001 edge | 65500 deny
- 65000: VNet to VNet 65001: LB in, internet out 65500: deny all Custom: 100-4096
Azure Firewall vs WAF DA
Azure Firewall
- Network and FQDN egress
- DNAT, network, application
- IDPS on Premium
WAF
- Inbound HTTP inspection
- SQL injection and XSS
- Gateway or Front Door
- Egress vs web requests
Azure SQL Security DA
- TDE Encrypts at rest, default
- TDE with CMK Key Vault wraps key
- Always Encrypted Client-side column encryption
- Dynamic data masking Hides values in results
- Row-level security Filters rows per user
- Ledger Tamper-evident change proof
- Server firewall rule Applies to all databases
- SQL auditing Storage, Log Analytics, Event Hubs
- Entra-only auth Blocks SQL logins
- Defender for SQL Vulnerability assessment plus alerts
Private Endpoint vs Service Endpoint DA
Private endpoint
- Private IP in VNet
- Reachable from on-premises
- Needs private DNS
Service endpoint
- Source stays public IP
- Subnet scoped
- No cost, no DNS
- Bring service in vs route out
NSG Default Rules DA
- Custom priority 100 to 4096
- Lower number Processed first, wins
- AllowVNetInBound Priority 65000
- AllowAzureLoadBalancerInBound Priority 65001
- DenyAllInbound Priority 65500
- AllowVnetOutBound Priority 65000
- AllowInternetOutBound Priority 65001
- DenyAllOutBound Priority 65500
- Stateful Return traffic auto-allowed
- ASG Groups NICs, not addresses
Firewall Rule Order DA
- Threat intelligence Runs before all rules
- DNAT rules Inbound translation, evaluated first
- Network rules Layer 3-4, after DNAT
- Application rules FQDN based, evaluated last
- Terminating First match stops processing
- Parent policy Beats child policy priority
- Premium SKU TLS inspection, IDPS, URL filtering
- IDPS Alert Runs parallel, logs only
- IDPS Alert and Deny Inline, drops matching flow
- Default action Deny anything unmatched
Private Access Options DA
- Private endpoint Private IP inside VNet
- Private Link service Publish your own service
- Service endpoint Subnet keeps public IP
- Private DNS zone Resolves FQDN to private
- Connection approval Automatic or manual request
- Entra Private Access Replaces VPN for apps
- Quick Access Primary FQDN and IP set
- Network Watcher Shows effective security rules
Security Admin Rules DA
- Owner Azure Virtual Network Manager
- Scope Applied at virtual network
- Order Evaluated before NSG rules
- Allow NSG rules still evaluated
- Always Allow Terminates, bypasses NSG
- Deny Terminates, drops traffic
- Priority 1 to 4096
- One config per region Use multiple rule collections
AI Security Stack CO
- Discover -> Identity -> Guardrail -> Detect
- Discover: Purview DSPM Identity: Entra Agent ID Guardrail: Prompt Shields Detect: Defender for AI
Agent ID vs Managed Identity CO
Entra Agent ID
- Identity for AI agents
- Blueprints create agents
- Conditional Access applies
Managed identity
- Identity for Azure resources
- System or user assigned
- Gets RBAC roles
- Agent vs resource
AI Control Picker CO
- Copilot sees too much → Purview DSPM for AI
- SharePoint oversharing → Data risk assessment
- Agent needs an identity → Entra Agent ID
- Limit agent access → Conditional Access ( Target agents )
- Trace agent blast radius → Defender XDR
- Block jailbreak prompts → Prompt Shields ( Foundry guardrail )
- Cap model token spend → APIM token limit
- Alert on AI attacks → Defender for AI Services
- Watch AI posture → Data and AI dashboard
Security for AI CO
- Purview DSPM for AI Copilot and AI data risk
- Data risk assessment Finds SharePoint oversharing
- Restricted Content Discovery Hides sites from Copilot
- Agent runtime protection Copilot Studio security status
- Entra Agent ID Identities for AI agents
- Agent identity blueprint Template creating agent identities
- Conditional Access for agents Blocks or limits agents
- Risky Agents report Agent risk from ID Protection
- Defender XDR Agent blast radius analysis
- Microsoft 365 admin center Where agents are managed
AI Guardrails CO
- Foundry guardrails Filters for models and agents
- Prompt Shields Detects prompt injection attacks
- User prompt attack Direct jailbreak attempt
- Document attack Indirect injection in content
- Spotlighting Marks untrusted document text
- Intervention points Input, tool call, output
- Guardrail action Annotate or block
- Harm categories Hate, Sexual, Violence, Self-Harm
- Trimmed severity 0, 2, 4, 6
- Groundedness detection Flags unsupported model claims
APIM AI Gateway CO
- llm-token-limit Caps tokens per consumer
- llm-emit-token-metric Token metrics to Monitor
- llm-semantic-cache-lookup Reuses similar prompt completions
- llm-content-safety Moderates prompts inline
- Backend load balancer Round-robin, weighted, priority
- Circuit breaker Stops calling failing backend
- Managed identity auth Removes stored API keys
VM and Server Hardening CO
- Trusted launch Default for new Gen2
- Secure boot Only signed components boot
- vTPM Measures the boot chain
- Integrity monitoring Remote boot attestation health
- Encryption at host Encrypts on the host
- Azure Disk Encryption In-guest BitLocker or DM-Crypt
- Bastion RDP/SSH without public IP
- AzureBastionSubnet /26 or larger
- JIT VM access Opens ports on request
- Azure Arc Extends controls to hybrid
- Machine Configuration Enforces guest OS baselines
Containers and App Platform CO
- Defender for Containers Misconfiguration and runtime risk
- Registry scanning Container image vulnerabilities
- AKS hardening Cluster security recommendations
- App Service auth Built-in identity provider
- Access restrictions IP allow list for apps
- Functions network access Private endpoints, inbound rules
- Logic Apps Secure inputs and outputs
- WAF modes Detection or Prevention
- APIM policies Protect back-end APIs
Sentinel Retention PO
- 30 default | 730 interactive | 12 years total
- Analytics default 30 days Interactive max 730 days Basic and Auxiliary fixed 30 Total retention 12 years
Sentinel vs Defender XDR PO
Sentinel
- Cloud-native SIEM
- Any connector or syslog
- KQL and playbooks
Defender XDR
- Microsoft-first XDR
- Correlates Microsoft alerts
- Agent blast radius
- Any source vs Microsoft
Defender Plan Picker PO
- Need attack paths → Defender CSPM ( Paid plan )
- Need EDR only → Servers Plan 1
- Need JIT and FIM → Servers Plan 2
- Scan blobs for malware → Defender for Storage
- Database threat alerts → Defender for Databases
- Find hardcoded secrets → Defender CSPM ( Secrets scanning )
- Vault access threats → Defender for Key Vault
- Container runtime risk → Defender for Containers
- Unknown internet assets → Defender EASM
- Protect AWS and GCP → Multicloud connector
CSPM Plan Split PO
- Foundational CSPM Free posture plan
- MCSB Default free security standard
- Secure score Free, based on MCSB
- Asset inventory Free in both plans
- Defender CSPM Paid advanced posture
- Attack path analysis Defender CSPM only
- Cloud security explorer Defender CSPM only
- Agentless VM scanning Defender CSPM only
- AI security posture Defender CSPM only
- Defender EASM Finds unprotected internet assets
Foundational vs Defender CSPM PO
Foundational
- Free plan
- MCSB and secure score
- Recommendations and inventory
Defender CSPM
- Paid plan
- Attack paths and explorer
- Agentless and AI posture
- Score vs risk context
Defender Workload Plans PO
- Servers Plan 1 Defender for Endpoint integration
- Servers Plan 2 Adds JIT, agentless, FIM
- File integrity monitoring Plan 2 only
- Data ingestion benefit Plan 2 free allowance
- Defender for Storage Priced per storage account
- Malware scanning Billed per GB scanned
- Defender for Databases Azure database threat alerts
- Defender for AI Services Generative AI threat alerts
- Multicloud connectors Onboard AWS and GCP
Sentinel Setup PO
- Workspace Log Analytics is required
- Content hub Installs solutions and connectors
- Sentinel Reader Views data and incidents
- Sentinel Responder Adds incident management
- Sentinel Contributor Creates rules, manages content
- Playbook Operator Runs playbooks manually
- Logic App Contributor Creates and edits playbooks
- Automation Contributor Service role, not people
- Automation rule Orchestrates incident response
- Playbook Logic App workflow
Sentinel Data Collection PO
- Data collection rule Defines what AMA collects
- Syslog via AMA Lands in Syslog table
- CEF via AMA Lands in CommonSecurityLog
- Log forwarder Linux VM running rsyslog
- Windows Security events Collected through a DCR
- Custom log table Stores ingested custom data
- Analytics default 30 days interactive
- Analytics maximum 730 days interactive
- Basic and Auxiliary Fixed 30-day interactive
- Total retention Up to 12 years
Security Copilot PO
- SCU Security compute unit
- Provisioned capacity Baseline, billed hourly
- Overage capacity Absorbs usage spikes
- Capacity calculator Estimates SCUs needed
- Owner role Changes capacity units
- Plugins Extend Copilot data sources
- Workspace Where capacity is associated
Common Traps
Sign in vs act
- Entra ID authenticates ≠ Azure RBAC authorizes
Eligible vs active
- Eligible needs activation ≠ Active works immediately
Soft delete vs purge protection
- Soft delete always on ≠ Purge protection is optional
Allow vs Always Allow
- Allow still hits NSG ≠ Always Allow skips NSG
Private vs service endpoint
- Private endpoint gets private IP ≠ Service endpoint keeps public IP
Firewall vs WAF
- Firewall filters network egress ≠ WAF inspects HTTP requests
Servers Plan 1 vs 2
- Plan 1 is EDR ≠ Plan 2 adds JIT
Free vs paid posture
- Foundational gives secure score ≠ Defender CSPM gives attack paths
Detection vs prevention
- Detection only logs ≠ Prevention blocks the request
Masking vs encryption
- Masking hides query output ≠ Always Encrypted protects stored data
Last Minute
- 1. Storage, database, network is heaviest: 25-30%
- 2. Other three areas: 20-25% each
- 3. Pass = 700 of 1000
- 4. Entra authenticates; Azure RBAC authorizes
- 5. PIM eligible means activate first
- 6. Key Vault soft delete: 7-90 days
- 7. Purge protection is not default
- 8. NSG deny-all inbound is 65500
- 9. Firewall: DNAT, network, then application
- 10. Always Allow bypasses NSG rules
- 11. Private endpoint = private VNet IP
- 12. WAF stops HTTP attacks, not egress
- 13. JIT needs Defender Servers Plan 2
- 14. Attack paths need paid Defender CSPM
- 15. DSPM for AI finds Copilot oversharing
- 16. Prompt Shields block jailbreak and injection
- 17. Entra Agent ID identifies AI agents
- 18. Sentinel default retention: 30 days
Microsoft Certified: Cloud and AI Security Engineer Associate (SC-500)
- Practice Questions 116 questions Study Guide 49 topics Cheat Sheet 203 facts Flashcards 50 cards Free CE Podcast