Microsoft SC-100 Practice Test 23
Question 220
Lakeshore Group is reviewing its Zero Trust strategy assessment. An internal AI assistant can retrieve internal reference files beyond its intended knowledge scope. Evidence: its retrieval identity has broad data access and prompt-injection tests are missing. The architect must apply data-access and threat-model controls to an AI solution. Security must preserve an approved incident response and recovery path. Which limited deployment approach best reduces rollout risk while testing the control?
- Apply the benchmark mapping to a pilot subscription and assess reported noncompliance
- Pilot a deployment gate with one signed approved image and one altered or unsigned image
- Pilot the conditional rule with a protected test admin and a healthy and noncompliant device
- Test the AI assistant on bounded synthetic documents and known indirect prompt-injection samples
Explanation
Correct Answer
D:
Its retrieval identity has broad data access and prompt-injection tests are missing. An AI workload baseline should address model access, data exposure, prompt-injection boundaries and sensitive output handling. Test the AI assistant on bounded synthetic documents and known indirect prompt-injection samples.
Incorrect Answers
A:
Microsoft Cloud Security Benchmark mappings connect security requirements to assessable cloud control categories and implementation ownership. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
B:
Build provenance links a deployed artifact to trusted source, dependencies and a traceable controlled build pipeline. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
C:
Zero Trust checks identity, device and context for each access decision instead of relying on network location alone. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
Question 221
Lakeshore Group is reviewing its Zero Trust strategy assessment. An internal AI assistant can retrieve internal reference files beyond its intended knowledge scope. Evidence: its retrieval identity has broad data access and prompt-injection tests are missing. The architect must apply data-access and threat-model controls to an AI solution. Security must preserve an approved incident response and recovery path. Which target-state architectural control most directly meets the requirement?
- Adopt least-privilege retrieval access, content controls, monitoring, and adversarial evaluation aligned to cloud security benchmarks
- Require trusted image digests and supported artifact provenance verification before deployment
- Require Conditional Access evaluation of appropriate identity and device signals for administrative access
- Define a baseline initiative and documented owners for control gaps consistent with MCSB
Explanation
Correct Answer
A:
Its retrieval identity has broad data access and prompt-injection tests are missing. An AI workload baseline should address model access, data exposure, prompt-injection boundaries and sensitive output handling. Adopt least-privilege retrieval access, content controls, monitoring, and adversarial evaluation aligned to cloud security benchmarks.
Incorrect Answers
B:
Build provenance links a deployed artifact to trusted source, dependencies and a traceable controlled build pipeline. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
C:
Zero Trust checks identity, device and context for each access decision instead of relying on network location alone. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
D:
Microsoft Cloud Security Benchmark mappings connect security requirements to assessable cloud control categories and implementation ownership. The AI security baseline must address prompt injection, model and data access, tool permission, and unsafe output together; a single conventional perimeter control cannot cover these trust boundaries.
Question 222
Amberline Group is reviewing its privileged access hardening initiative. A helpdesk service account has administrative access to domain controllers and ordinary workstations. Evidence: the same credentials are reused across privileged and low-trust hosts. The architect must harden domain controllers and Tier 0 identity boundaries. Security must preserve an approved incident response and recovery path. Which investigation should the security architect perform first to validate the cause and dependencies?
- Inspect active and eligible role assignments plus existing PIM activation controls
- Inventory privileged group membership, domain controller logons, service account rights, and lateral paths
- Inspect diagnostic settings, workspace selection, retention and ingestion gaps
- Inspect privileged sign-in device compliance and the current administrative access path
Explanation
Correct Answer
B:
The same credentials are reused across privileged and low-trust hosts. Tier 0 controls isolate highly privileged Active Directory administration and directory controllers from ordinary workstation exposure. Inventory privileged group membership, domain controller logons, service account rights, and lateral paths.
Incorrect Answers
A:
Privileged Identity Management limits standing privilege by requiring eligible role activation and recording supported approval or audit events. Tier 0 identity security keeps domain controller administration and highly privileged accounts separate from general-purpose endpoints, preventing lower-tier compromise from inheriting directory control.
C:
Sentinel detection and hunting depend on ingesting the relevant event sources into an authorized Log Analytics workspace. Tier 0 identity security keeps domain controller administration and highly privileged accounts separate from general-purpose endpoints, preventing lower-tier compromise from inheriting directory control.
D:
A privileged-access workstation hardens the endpoint used by an administrator. It cannot, on its own, separate Tier 0 directory controller administration and privileges from lower-trust identities and workstation paths.
Question 223
Bluebird Group is reviewing its regulatory security governance initiative. A government agency must demonstrate that new storage resources use mandated encryption. The architect must use Azure Policy to prevent or discover prohibited resource states. The solution must not grant access beyond a named pilot group. Which acceptance test would provide the strongest evidence that the design works?
- Verify the mapped control assessment updates following correction while noting remaining manual requirements
- Confirm contractual records remain retained and permissible records follow approved disposition rules
- Verify the prohibited configuration is denied or flagged according to the assigned effect
- Verify the allowed job exchanges OIDC for an Azure token while an untrusted branch cannot
Explanation
Correct Answer
C:
A government agency must demonstrate that new storage resources use mandated encryption. Azure Policy can audit or enforce supported resource properties against centrally assigned governance requirements. Verify the prohibited configuration is denied or flagged according to the assigned effect.
Incorrect Answers
A:
Regulatory compliance views map relevant cloud configuration evidence to selected standards, not legal compliance guarantees. Azure Policy assignments must audit or deny the prohibited resource configuration at the intended scope; documentation or recommendation scores are not enforcement.
B:
A retention rule governs how long business records are preserved and when they may be disposed of. It cannot evaluate Azure resource settings against an assigned cloud-policy baseline; that is Azure Policy's configuration-control role.
D:
Workload identity federation avoids client secrets in a deployment pipeline, but it cannot prevent a prohibited Azure resource configuration. Azure Policy assignments with appropriate deny or audit effects address resource-state governance.
Question 224
Granite Group is reviewing its security reference architecture migration. A software vendor wants temporary production troubleshooting without persistent tenant-wide privileges. The architect must apply just-enough and just-in-time access to operational roles. Production disruption must be avoided and the rollout must be reversible. Which ongoing signal would most directly detect renewed failure of this control?
- Review access decisions that bypass contextual checks and monitor sign-in risk trends
- Watch denied east-west traffic and high-risk privilege or egress changes
- Review cross-subscription role sprawl and unauthorized platform configuration changes
- Audit role assignment creep, stale guest access, and repeated unnecessary elevations
Explanation
Correct Answer
D:
A software vendor wants temporary production troubleshooting without persistent tenant-wide privileges. Least privilege grants identities only required permissions and reduces the impact of credential compromise. Audit role assignment creep, stale guest access, and repeated unnecessary elevations.
Incorrect Answers
A:
Zero Trust checks identity, device and context for each access decision instead of relying on network location alone. Just-enough and just-in-time roles constrain both permission breadth and duration; an administrative workstation or audit trail does not replace least-privilege role design.
B:
Assume-breach design contains compromise through segmentation, monitoring and restricted east-west movement. Just-enough and just-in-time roles constrain both permission breadth and duration; an administrative workstation or audit trail does not replace least-privilege role design.
C:
Landing zones isolate platform operations and application workloads with defined RBAC, policy and network boundaries. Just-enough and just-in-time roles constrain both permission breadth and duration; an administrative workstation or audit trail does not replace least-privilege role design.
Question 225
Riverbend Group is reviewing its multi-tenant access assurance project. A release pipeline in a partner repo needs short-lived deployment access without sharing client secrets. The architect must avoid stored cloud credentials in a build pipeline. Existing audit and retention obligations must remain intact. Which ongoing signal would most directly detect renewed failure of this control?
- Monitor federated credential subject drift and unexpected service-principal grants
- Review managed identity role grants and new secret usage in application configurations
- Monitor strength evaluation failures and changes to privileged group targeting
- Monitor agent identity inventory, permission drift, and unused agents
Explanation
Correct Answer
A:
A release pipeline in a partner repo needs short-lived deployment access without sharing client secrets. CI/CD workload federation exchanges trusted short-lived identity assertions instead of storing static cloud client secrets. Monitor federated credential subject drift and unexpected service-principal grants.
Incorrect Answers
B:
A runtime managed identity is attached to an eligible hosted workload; it does not automatically eliminate client secrets used by an external repository's deployment job. Trusted CI/CD workload federation provides short-lived credentials for that build identity.
C:
Authentication strengths require approved sign-in method combinations for selected Conditional Access scenarios. Federated CI/CD credentials exchange trusted short-lived assertions for cloud tokens; keeping a static client secret in a secured store still leaves rotation and leakage risk.
D:
Agent identities need attributable ownership and constrained application permissions to limit autonomous tool access. Federated CI/CD credentials exchange trusted short-lived assertions for cloud tokens; keeping a static client secret in a secured store still leaves rotation and leakage risk.
Question 226
Keystone Group is reviewing its Azure application infrastructure review. A clinical document analysis service must keep inference access inside a private application network. The architect must restrict AI service invocation to intended network and identity paths. Architecture owners require evidence for both allowed and denied actions. Which limited deployment approach best reduces rollout risk while testing the control?
- Pilot scanning and deployment validation on both an approved and deliberately outdated test image
- Pilot requests from an approved private subnet and a blocked public client
- Pilot the encryption configuration on an eligible nonproduction managed disk
- Pilot one recommended secure setting on a representative nonproduction resource
Explanation
Correct Answer
B:
A clinical document analysis service must keep inference access inside a private application network. Private endpoints and service network restrictions can limit supported Azure AI endpoints to approved private connectivity. Pilot requests from an approved private subnet and a blocked public client.
Incorrect Answers
A:
Image scanning identifies vulnerable dependencies, while admission policy can block disallowed images at deployment. Network-restricted AI access needs approved private endpoint routing and service authorization; a model-risk review does not prevent unauthorized endpoint connections.
C:
Disk encryption and key governance protect VM volume data and control who may use or rotate the required cryptographic keys. Network-restricted AI access needs approved private endpoint routing and service authorization; a model-risk review does not prevent unauthorized endpoint connections.
D:
Cloud Secure Score highlights posture improvements, but actual remediation should prioritize exposed business risks. Network-restricted AI access needs approved private endpoint routing and service authorization; a model-risk review does not prevent unauthorized endpoint connections.
Question 227
Silverline Group is reviewing its application security lifecycle modernization. A deployment service accepts binaries without a verifiable relationship to the approved source commit. Evidence: artifacts can be replaced between CI completion and production release. The architect must bind released application artifacts to trusted build and review provenance. Architecture owners require evidence for both allowed and denied actions. Which target-state architectural control most directly meets the requirement?
- Create a prioritized threat model with corresponding control requirements and owners
- Require supported code/secret scanning, secret revocation, and reviewed pull-request security checks
- Use supported provenance attestations and immutable artifact digests in release approval
- Apply APIM validate-jwt and rate-limiting policies for eligible API operations
Explanation
Correct Answer
C:
Artifacts can be replaced between CI completion and production release. Signed or verifiable deployment attestations establish an artifact’s build identity and release provenance. Use supported provenance attestations and immutable artifact digests in release approval.
Incorrect Answers
A:
STRIDE systematically considers spoofing, tampering, repudiation, information disclosure, denial of service and elevation of privilege. Artifact provenance requires matching the production binary digest to an approved attested build; threat models and prerelease scans cannot prove the exact object actually deployed.
B:
Code and secret scanning can reject risky changes before deployment when findings are wired into enforced pipeline gates. Artifact provenance requires matching the production binary digest to an approved attested build; threat models and prerelease scans cannot prove the exact object actually deployed.
D:
API Management JWT validation establishes caller authorization for an API request, not whether a released binary came from reviewed source in a trusted build. Signed build provenance and deployment digest checks address artifact integrity.
Question 228
Stonebridge Group is reviewing its secure application architecture assessment. A banking portal needs a specific checkout field excluded from inspection without turning WAF off. The architect must tune web application protection without disabling its entire rule set. Production disruption must be avoided and the rollout must be reversible. Which acceptance test would provide the strongest evidence that the design works?
- Verify authorized crypto operations succeed while private key material remains protected under the selected model
- Verify only authorized correctly scoped tokens pass and excess requests receive the expected throttle
- Verify the test alert blocks promotion and any exposed real credential is revoked and replaced
- Confirm legitimate transactions succeed while attacks in other paths remain blocked
Explanation
Correct Answer
D:
A banking portal needs a specific checkout field excluded from inspection without turning WAF off. WAF tuning should scope justified exclusions or custom rules without disabling unrelated protections. Confirm legitimate transactions succeed while attacks in other paths remain blocked.
Incorrect Answers
A:
Key ownership design separates cryptographic material, privileged operations, rotation responsibilities and application access. A WAF exclusion should target the demonstrated false-positive path or rule while leaving unrelated signatures active; disabling protection wholesale would recreate the exposure.
B:
API Management can reject unauthorized callers through JWT validation. It cannot fix a WAF managed rule that is blocking legitimate HTTP traffic; scope a tested WAF exclusion or custom rule without disabling unrelated protections.
C:
Code and secret scanning can reject risky changes before deployment when findings are wired into enforced pipeline gates. A WAF exclusion should target the demonstrated false-positive path or rule while leaving unrelated signatures active; disabling protection wholesale would recreate the exposure.
Question 229
Summit Group is reviewing its risk-driven exposure management program. A data-center migration needs unified inventory for legacy servers before cloud relocation. The architect must extend Azure governance and supported security management to hybrid assets. Security must preserve an approved incident response and recovery path. Which acceptance test would provide the strongest evidence that the design works?
- Verify the machine appears in Azure and receives intended supported policy and security configuration
- Verify the change is recognized by the assessment without breaking intended workloads
- Verify the database surfaces expected security findings and security alert routing is functional
- Confirm supported Azure, AWS, and GCP assets surface appropriate posture recommendations
Explanation
Correct Answer
A:
A data-center migration needs unified inventory for legacy servers before cloud relocation. Azure Arc extends supported Azure governance and inventory capabilities to enrolled non-Azure servers and platforms. Verify the machine appears in Azure and receives intended supported policy and security configuration.
Incorrect Answers
B:
Cloud Secure Score highlights posture improvements, but actual remediation should prioritize exposed business risks. Azure Arc enrollment brings supported non-Azure machines into inventory and policy management; configuring a cloud-only workload does not govern the hybrid assets in the requirement.
C:
Database Defender plans surface supported vulnerabilities or anomalous activity when coverage is correctly enabled. Azure Arc enrollment brings supported non-Azure machines into inventory and policy management; configuring a cloud-only workload does not govern the hybrid assets in the requirement.
D:
Defender for Cloud assesses supported cloud assets against security recommendations when connectors and appropriate plans are enabled. Azure Arc enrollment brings supported non-Azure machines into inventory and policy management; configuring a cloud-only workload does not govern the hybrid assets in the requirement.
Question 230
Solstice Group is reviewing its application security lifecycle modernization. A serverless function stores a shared access key to write messages into a queue. Evidence: the app runs on Azure compute and the destination supports identity-based authorization. The architect must authenticate the application with a managed identity and only required data-plane permissions. The design must support controlled testing before tenant-wide enforcement. Which acceptance test would provide the strongest evidence that the design works?
- Verify the test alert blocks promotion and any exposed real credential is revoked and replaced
- Verify message send succeeds while unauthorized queue receive and other storage operations fail
- Verify only authorized correctly scoped tokens pass and excess requests receive the expected throttle
- Verify production deployment accepts only an artifact matching the approved digest and provenance
Explanation
Correct Answer
B:
The app runs on Azure compute and the destination supports identity-based authorization. Workload identity federation avoids long-lived secrets while Azure RBAC constrains the actions performed by that identity. Verify message send succeeds while unauthorized queue receive and other storage operations fail.
Incorrect Answers
A:
Code and secret scanning can reject risky changes before deployment when findings are wired into enforced pipeline gates. Workload federation or managed identity issues scoped tokens without embedded credentials, while Azure data-plane RBAC grants only required operations; network filtering by itself does not authorize storage actions.
C:
API Management can validate supported OAuth/JWT access tokens before forwarding authorized requests to backend APIs. Workload federation or managed identity issues scoped tokens without embedded credentials, while Azure data-plane RBAC grants only required operations; network filtering by itself does not authorize storage actions.
D:
Signed or verifiable deployment attestations establish an artifact’s build identity and release provenance. Workload federation or managed identity issues scoped tokens without embedded credentials, while Azure data-plane RBAC grants only required operations; network filtering by itself does not authorize storage actions.