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?”

What you will learn
Make a defensible incident closure decision.
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
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
| Decision | Investigation question |
|---|---|
| True positive | Was meaningful suspicious or malicious activity confirmed? |
| Benign positive | Was the detection accurate but the activity legitimate? |
| False positive | Was the detection itself misleading or incorrect? |
| Unresolved / escalate | Is 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
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
| Check | What to record |
|---|---|
| Findings | Confirmed events and the strongest supporting evidence |
| Timeline | First known activity, major events and latest relevant activity |
| Scope | Affected identities, hosts, resources and known limitations |
| Classification | Chosen outcome and evidence-based reason |
| Actions | Containment, remediation, verification and owners |
| Follow-up | Monitoring, detection tuning and residual risks |
| Case record | Clear 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
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 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 ascThis 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
Illustrative only: do not copy claims of completed remediation unless they were actually verified in your incident.
Example closure note — benign activity
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.
- Would you close the incident now? Explain why.
- What additional evidence must be checked?
- Which entities belong in the final scope?
- What would make a true-positive classification defensible?
- 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.
