Agent Foskett Academy • Microsoft Sentinel • Module 4 • Lesson 44

Lesson 44 — Closing a Microsoft Sentinel Incident Properly

The alerts have stopped.

The incident looks quiet.

Someone asks: “Can we close it?”

Agent Foskett has one more question: “What evidence supports that decision?”

Closure is an investigation decision — not a way to make the dashboard green.
Agent Foskett Microsoft Sentinel Academy incident closure investigation
What you will learn

Make a defensible incident closure decision.

✓ Validate findings and scope
✓ Select the right classification
✓ Verify actions and ownership
✓ Document closure and follow-up

Learning objectives

  • Identify when an incident is ready for closure.
  • Distinguish true positives, benign positives and false positives.
  • Explain why alert resolution does not prove containment.
  • Record evidence, response actions and residual risks.
  • Write a clear closure comment and handover.
  • Recognise when an incident should remain open or be escalated.

Closing is not deleting

Closing an incident records the investigation's current disposition. It does not erase the underlying telemetry or automatically undo attacker activity.

Status is a workflow state, not a security guarantee.

Before you click Close

INCIDENT READY FOR CLOSURE? │ ├─ What actually happened? ├─ Which entities were affected? ├─ What evidence supports the conclusion? ├─ Were response actions completed and verified? ├─ Are risks or follow-up tasks still outstanding? └─ Can another analyst reproduce the decision? │ ▼ CLASSIFY → DOCUMENT → CLOSE (when justified)

If material questions remain unanswered, closure may be premature. Follow your organisation's escalation and case-management process.

True positive

The detection identified the behaviour it was designed to find and the activity warrants the relevant security interpretation. A true positive may require containment, remediation and further scoping before closure.

Benign positive

The detection fired on genuine observed activity, but investigation established an authorised or otherwise benign explanation. Document that explanation and consider whether detection tuning is warranted.

False positive

The detection did not correctly identify the behaviour of concern, such as when rule logic or data interpretation produced a misleading result. Preserve enough detail to support rule review.

Undetermined is not “benign”

If evidence is insufficient to establish the outcome, do not relabel the incident as harmless merely to close the queue. Escalate or follow the documented process for unresolved cases.

Classification and reason: make the distinction

DecisionInvestigation question
True positiveWas meaningful suspicious or malicious activity confirmed?
Benign positiveWas the detection accurate but the activity legitimate?
False positiveWas the detection itself misleading or incorrect?
Unresolved / escalateIs the available evidence inadequate for a defensible conclusion?

Exact classification labels and closure options can vary between Sentinel experiences and the Defender portal. Use the options presented by your environment and your SOC's procedures.

Case study — the incident closed too early

02:08 Suspicious mailbox rule alert 02:14 PowerShell execution alert 02:17 Suspicious network connection 02:20 Automation marks incident closed 02:31 Account still active; new suspicious sign-in DASHBOARD: CLOSED RISK: NOT RESOLVED

The status changed. The evidence did not. A closure rule that ignores related activity can hide an active threat.

Verify response actions

Do not assume “isolation requested” means “device isolated.” Check the recorded action result and relevant telemetry. Likewise, confirm whether account, session or network controls actually took effect.

Check incident scope

Revisit users, devices, IP addresses, applications and related incidents. The original alert may have named only one entity even when the investigation found more.

Closing checklist

CheckWhat to record
FindingsConfirmed events and the strongest supporting evidence
TimelineFirst known activity, major events and latest relevant activity
ScopeAffected identities, hosts, resources and known limitations
ClassificationChosen outcome and evidence-based reason
ActionsContainment, remediation, verification and owners
Follow-upMonitoring, detection tuning and residual risks
Case recordClear summary, evidence links and decision author

What about automation?

Automation can update incidents and reduce repetitive work, but broad closure conditions may suppress important cases. Review the automation rule's conditions, audit trail and impact before treating automatic closure as proof of benign activity.

Closure and tuning are different

Closing an incident records the case outcome. Tuning an analytics rule addresses future detection quality. An incident may be closed while a separate, tracked tuning task remains open.

KQL — check activity after the last alert

Validate post-alert identity activity
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
SigninLogs
| where UserPrincipalName =~ "alex@contoso.com"
| where TimeGenerated between (
    datetime(2026-10-07 02:17:00) ..
    datetime(2026-10-07 04:00:00)
)
| project TimeGenerated, UserPrincipalName, IPAddress,
          AppDisplayName, ResultType, ConditionalAccessStatus
| order by TimeGenerated asc

This query checks for sign-in activity after the last known alert in our illustrative case. It cannot, by itself, prove that an attacker has been removed or that every persistence mechanism is gone.

Beware the “no results” trap

No sign-ins in one queried table and window means exactly that. It does not prove that no other suspicious activity occurred. Consider data coverage, ingestion delays and other relevant sources.

Know when to escalate

Keep the case open or escalate when containment is unverified, affected assets remain unknown, suspicious activity continues or the business risk exceeds the analyst's authority.

Example closure note — confirmed incident

Classification: True positive Summary: Unauthorised mailbox rule activity confirmed for alex@contoso.com. Evidence: Sign-in event 01:57 UTC; mailbox audit activity 02:08 UTC; linked incident alerts. Scope: Account reviewed; related entities checked. Actions: Session revocation and account remediation completed; outcomes verified by response team. Residual risk: Additional monitoring assigned. Follow-up: Review conditional access and detection logic. Decision: Closed under SOC procedure after validation.

Illustrative only: do not copy claims of completed remediation unless they were actually verified in your incident.

Example closure note — benign activity

Classification: Benign positive Summary: Scheduled administrative activity matched the analytics rule. Evidence: Approved change record and matching administrator, device and event timestamps. Actions: No containment required. Follow-up: Detection tuning ticket assigned. Decision: Closed with documented justification.

The reason matters. “Known good” without supporting context is not an adequate explanation.

Incident activity history

Review case comments and recorded actions before closure. In the Defender portal, incident activity provides useful context about manual and automated changes to the case.

Ownership after closure

Closure should not silently abandon outstanding work. Assign owners and tracking references for any residual monitoring, engineering fixes or lessons learned.

Agent Foskett exercise

A high-severity incident includes suspicious sign-in and mailbox-rule alerts. The mailbox rule was removed, but nobody has confirmed whether active sessions were revoked. Another analyst found a second affected user.

  1. Would you close the incident now? Explain why.
  2. What additional evidence must be checked?
  3. Which entities belong in the final scope?
  4. What would make a true-positive classification defensible?
  5. Write a closure-ready summary only after the missing actions are verified.

Common mistakes

  • Closing because no new alert appeared.
  • Assuming containment was successful without verification.
  • Using false positive as a synonym for benign activity.
  • Ignoring related entities and incidents.
  • Leaving no evidence-based closure reason.
  • Allowing automation to hide unresolved activity.

Agent Foskett takeaway

Close the case when the evidence supports the outcome — not when the dashboard asks for a smaller number.

The logs already knew. Your closure notes should show what they told you.

Lesson summary
Closing a Microsoft Sentinel incident properly means validating evidence and scope, choosing a defensible classification, verifying response actions and documenting any remaining risk or follow-up.
Sentinel Academy Home

Related Agent Foskett learning

Use your incident timeline, evidence and analyst notes to support the final decision.