Home / Enterprise Use Cases
๐Ÿ›ก๏ธ

Microsoft Security: Enterprise Use Cases

End-to-end scenarios showing how Defender XDR, Sentinel, Security Copilot, and Purview work together under a Zero Trust framework. from detection to remediation, with KQL queries, AI prompts, playbooks, and lessons learned.

The Microsoft Unified Security Operations Model

Modern enterprise security requires an integrated platform where machines detect, AI triages, humans decide, and automation responds. These use cases demonstrate this model in action, showing specific tools, queries, and workflows at each stage.

Zero Trust Security Operations Workflow

1. Detect
Defender XDR auto-correlates
alerts across 9 products
2. Enrich
Sentinel adds 3rd-party data,
custom KQL, threat intel
3. Investigate
Security Copilot summarises,
generates KQL, enriches IOCs
4. Respond
XDR AIR + Sentinel playbooks
auto-contain & remediate
5. Learn
Post-incident review, new
detections, posture hardening

Solutions Referenced in These Use Cases

Defender XDR Defender for Endpoint (MDE) Defender for Office 365 (MDO) Defender for Identity (MDI) Defender for Cloud Apps (MDA) Microsoft Sentinel Entra ID Protection Security Copilot Purview DLP Defender for Cloud (MDC)

Security Frameworks & Acronyms Used

Each use case is tagged with industry-standard security frameworks. Here is what they mean and why they matter.

๐Ÿ”’ Zero Trust Security Model

Zero Trust is a security strategy that assumes no user, device, or network is trustworthy by default. even inside the corporate network. Instead of trusting anything behind the firewall, every access request is fully authenticated, authorised, and encrypted before granting access. It is built on three core principles:

๐Ÿ” Verify Explicitly
Always authenticate and authorise based on all available data points: user identity, location, device health, service, data classification, and anomalies.
๐Ÿ” Least Privilege Access
Limit user access to only what they need, only when they need it. Use just-in-time and just-enough-access (JIT/JEA), risk-based adaptive policies, and data protection.
โš ๏ธ Assume Breach
Operate as if attackers are already inside. Minimise blast radius and segment access. Verify end-to-end encryption, use analytics to detect threats, and improve defences continuously.

๐Ÿ“‹ NIST Cybersecurity Framework (CSF 2.0)

The NIST Cybersecurity Framework, created by the U.S. National Institute of Standards and Technology, is the most widely adopted framework for organising cybersecurity activities. It defines six core functions. When you see badges like NIST: Detect · Respond on a use case, it tells you which functions that scenario exercises:

CodeFunctionWhat It MeansExample Activities
GVGovernEstablish and monitor the organisation’s cybersecurity risk management strategy and policiesRisk assessments, security policies, roles & responsibilities, compliance requirements
IDIdentifyUnderstand your assets, risks, and business context so you know what to protectAsset inventory, risk assessment, business impact analysis, supply chain risk
PRProtectImplement safeguards to prevent or limit the impact of a security eventMFA, encryption, endpoint protection, access controls, security training
DEDetectDiscover cybersecurity events in a timely manner using monitoring and analyticsSIEM alerts, XDR correlation, anomaly detection, threat hunting with KQL
RSRespondTake action when a cybersecurity incident is detected to contain the damageIncident response, containment, forensics, communication, playbook execution
RCRecoverRestore capabilities and services after an incident, and apply lessons learnedSystem restoration, backup recovery, post-incident review, process improvements

๐ŸŽฏ STRIDE Threat Model

STRIDE is a threat classification model developed by Microsoft to help security teams categorise what an attacker is trying to achieve. Each letter represents a category of threat. When you see a badge like STRIDE: Spoofing, it identifies the primary threat types in that scenario:

LetterThreatWhat the Attacker DoesExample
SSpoofingPretends to be someone or something they are notPhishing email impersonating a CFO, stolen credentials, forged authentication tokens
TTamperingModifies data or code without authorisationRansomware encrypting files, modifying firewall rules, altering backup agents
RRepudiationDenies having performed an action (no audit trail)Deleting logs to cover tracks, disabling auditing, clearing browser history
IInformation DisclosureAccesses data they should not be able to seeReading confidential emails, exfiltrating design files, dumping a database
DDenial of Service (DoS)Makes a system or service unavailableRansomware making servers inaccessible, disabling backup services, cryptomining consuming resources
EElevation of PrivilegeGains higher access than authorisedExploiting a vulnerability to gain admin rights, Golden Ticket attack, lateral movement to domain controller

๐Ÿ“– Microsoft Security Product Acronyms

These acronyms appear throughout the use cases. Here is the full name of each product and what it does:

AcronymFull NameWhat It Does
XDRExtended Detection & ResponseUnified platform that correlates alerts from all Defender products into single incidents
MDEMicrosoft Defender for EndpointProtects laptops, desktops, and servers. detects malware, ransomware, and suspicious behaviour
MDOMicrosoft Defender for Office 365Protects email and collaboration tools. blocks phishing, malicious attachments, and unsafe links
MDIMicrosoft Defender for IdentityMonitors Active Directory for identity-based attacks like credential theft and lateral movement
MDAMicrosoft Defender for Cloud AppsMonitors cloud application usage, detects shadow IT, and enforces session-level controls
MDCMicrosoft Defender for CloudProtects cloud workloads (VMs, containers, databases) across Azure, AWS, and GCP
AIRAutomated Investigation & ResponseXDR’s built-in automation that investigates alerts and takes remediation actions without human intervention
KQLKusto Query LanguageThe query language used in Sentinel and Defender XDR to search and analyse security data
SCUSecurity Compute UnitThe billing unit for Security Copilot. each SCU provides a fixed amount of AI processing capacity
DLPData Loss PreventionPolicies that detect and block sensitive data from being shared outside the organisation
IRMInsider Risk ManagementPurview feature that detects risky user behaviour such as data theft by departing employees
UEBAUser and Entity Behaviour AnalyticsSentinel feature that builds behavioural baselines and detects anomalies for users and devices
SOARSecurity Orchestration, Automation & ResponseAutomated playbooks (Logic Apps) that run response actions when incidents are detected
Jump to use case 1. Ransomware Attack & Recovery 2. Business Email Compromise 3. Insider Data Exfiltration 4. Cloud-Native SOC Transformation

Cross-Scenario Attack and Response Timeline

A consolidated view of how attacks unfold across all four enterprise scenarios. from initial compromise to full recovery and lessons learned.

🔒 Ransomware Attack (UC1)

Day 0. Initial Access
VPN Credential Compromise
Attacker uses stolen VPN credentials from dark web to access hospital network perimeter.
T1078 Valid Accounts
Day 1-3. Discovery
Internal Reconnaissance
Lateral movement across 23 systems. Active Directory enumeration, backup server identification.
T1018 Remote System Discovery
Day 4-7. Preparation
Backup Tampering and Staging
Shadow copies deleted. Backup agent disabled on 4 servers. Encryption payload staged via scheduled tasks.
T1490 Inhibit System Recovery
Day 9. T+00:00
XDR Correlation Triggers
Defender XDR correlates 23 alerts across MDE, MDI, and Sentinel into a single Critical incident.
Defender XDRSentinel
Day 9. T+00:04
Automated Containment
AIR isolates 23 endpoints. Copilot generates executive briefing. Sentinel playbook blocks lateral IPs.
AIRCopilot
Day 9. T+04:00
Full Recovery
Systems restored from offline backups. $14.8M cost avoided. Zero patient data compromised.
Recovery

📧 Business Email Compromise (UC2)

Hour 0. Initial Access
CFO Credential Phish
Spear-phishing email with fake DocuSign link targets CFO. AiTM proxy captures session token.
T1566.002 Spear-phishing Link
Hour 1. Persistence
Inbox Rule Creation
Attacker creates hidden inbox rules forwarding finance emails. Registers malicious OAuth app.
T1564.008 Email Rules
Hour 3. Execution
Wire Transfer Request
Fraudulent $2.4M wire transfer request sent to accounts payable from CFO's compromised account.
T1534 BEC
Hour 3.5. Detection
5-Signal Correlation
MDO detects impossible travel. MDA flags risky OAuth. Sentinel UEBA triggers anomaly alert.
MDOMDASentinel
Hour 4. Response
Account Lockdown
Session revoked, inbox rules deleted, OAuth app consent revoked, wire transfer frozen by bank.
CopilotApp Governance
Hour 6. Recovery
Full Remediation
$2.4M wire transfer reversed. MFA enforced on all executive accounts. Anti-phishing policies hardened.
Recovery

🕵 Insider Data Exfiltration (UC3)

Day 0. Trigger
Resignation Submitted
Senior engineer submits resignation. HR system triggers Insider Risk Management policy.
IRM Trigger
Day 1-3. Collection
Mass SharePoint Downloads
2,400 files downloaded from chip design SharePoint library. 340% above 90-day baseline.
T1530 Data from Cloud Storage
Day 4-5. Staging
USB and Personal Email
Files copied to personal USB drive. 180 files emailed to personal Gmail account.
T1052 Exfiltration over Physical
Day 5. Detection
Adaptive Protection Triggers
Purview DLP blocks USB copy. Insider Risk score reaches Critical. Sentinel correlation alert fires.
PurviewDLP
Day 6. Response
Legal and HR Investigation
Copilot quantifies exposure: 2,400 files, $50M+ IP value. eDiscovery case created. Legal hold applied.
CopiloteDiscovery
Day 8. Resolution
Containment Complete
All exfiltrated data identified. Employee access revoked. Forensic evidence package delivered to legal.
Recovery

🛠 SOC Transformation (UC4)

Month 1-2. Phase 1
Foundation
Deploy Sentinel workspace. Connect Entra ID, M365, and Defender XDR data connectors.
SentinelData Connectors
Month 3-4. Phase 2
Detection Engineering
Migrate 120 Splunk detection rules to KQL analytics rules. Build 15 automated playbooks.
KQLPlaybooks
Month 5-6. Phase 3
Automation and AI
Deploy Security Copilot for Tier 1 triage. Automate 60% of routine alerts. Build promptbooks.
CopilotAutomation
Month 7. Phase 4
Optimization and Decommission
Decommission Splunk. 82% faster MTTD. $2M annual savings. SOC team upskilled on KQL and Copilot.
ROI: $2M/year82% faster
💾
Simulation Scripts Available
Download PowerShell scripts to simulate and remediate these scenarios in your own lab environment.
Download Script
๐Ÿ”’

Use Case 1: Enterprise Ransomware Attack & Recovery

Zero Trust: Assume Breach NIST: Protect ยท Detect ยท Respond ยท Recover STRIDE: Tampering ยท Denial of Service ยท Elevation of Privilege

๐Ÿ“‹ Enterprise Scenario

Organisation: Regional hospital network ยท 8,000 endpoints ยท 12 facilities ยท 3,500 employees ยท HIPAA-regulated
Attack vector: Compromised VPN credential (no MFA enforced) โ†’ 9 days of undetected lateral movement โ†’ backup agent tampering โ†’ encryption at 2:00 AM Saturday
Potential impact: $15M+ recovery costs ยท 3–6 week operational disruption ยท patient safety risks ยท HIPAA breach notification

1 Phase 1: Detection. How the Platform Catches the Attack

โš”๏ธ Defender XDR: Automated Alert Correlation

Defender XDR automatically correlates 23 individual alerts from 5 products into 1 unified incident:

  • Entra ID Protection: Risky sign-in from anonymous IP address (VPN compromise) โ†’ CA policy should have blocked this
  • MDI: Kerberoasting detected โ†’ service account TGS requests from non-service workstation
  • MDE: Encoded PowerShell execution on finance workstation (first-ever PS execution on this device)
  • MDE: LSASS credential dumping attempt blocked by ASR rule
  • MDE: Lateral movement via admin shares to 14 servers

๐Ÿ“ก Sentinel: Custom KQL Detection

A custom Sentinel analytics rule detects backup agent service disruption. a critical early warning that traditional tools miss:

// Sentinel Analytics Rule: Backup Agent Service Stops
// Severity: High | Run every: 5 min | Look back: 10 min
Syslog
| where Facility == "daemon" and SeverityLevel == "err"
| where ProcessName in ("veeamservice", "VeeamBackupSvc", "wbengine")
| where SyslogMessage has_any ("stopped", "terminated", "disabled")
| summarize StoppedCount = dcount(Computer), 
            AffectedServers = make_set(Computer) by bin(TimeGenerated, 5m)
| where StoppedCount >= 3 // 3+ backup agents stopped = suspicious
| extend AlertTitle = strcat("CRITICAL: ", StoppedCount, 
         " backup agents stopped in 5 minutes")
| project TimeGenerated, AlertTitle, StoppedCount, AffectedServers

Entity mapping: Host โ†’ AffectedServers | MITRE: T1490 Inhibit System Recovery | Auto-response: Triggers playbook to restart services

2 Phase 2: AI-Accelerated Investigation with Security Copilot

๐Ÿ›ก๏ธ Copilot Incident Triage Promptbook (runs in 90 seconds)

The SOC has a custom Ransomware Triage Promptbook that runs automatically when the unified incident is opened:

Prompt 1: Attack Summary

“Summarise incident INC-2026-4891 including all affected entities, the attack timeline, MITRE ATT&CK techniques used, and the current blast radius.”

Copilot output: Generates a structured narrative identifying the initial access (compromised VPN), 14 affected devices, 3 compromised accounts, and 7 MITRE techniques across the kill chain.

Prompt 2: Lateral Movement Scope

“Generate a KQL query showing all devices the compromised user john.doe@contoso.com accessed via SMB admin shares in the last 14 days, including process details.”

// Copilot-generated KQL
DeviceLogonEvents
| where Timestamp > ago(14d)
| where AccountUpn =~ "john.doe@contoso.com"
| where LogonType in ("RemoteInteractive", "Network")
| where ActionType == "LogonSuccess"
| summarize FirstSeen=min(Timestamp), LastSeen=max(Timestamp),
    LogonCount=count() by DeviceName, RemoteIP, LogonType
| join kind=inner (
    DeviceProcessEvents
    | where InitiatingProcessAccountUpn =~ "john.doe@contoso.com"
    | where FileName in~ ("cmd.exe","powershell.exe","wmic.exe")
    | summarize Commands=make_set(ProcessCommandLine) by DeviceName
) on DeviceName
| sort by FirstSeen asc

Prompt 3: Backup Impact Assessment

“Which backup agents were disabled vs. still operational? Can we recover from the last clean backup?”

Copilot output: Cross-references Sentinel Syslog data with CMDB to report: 5 of 12 backup agents disabled. 7 agents operational with last clean backup at 01:30 AM (30 minutes before encryption started). Recovery feasible for all critical systems.

Prompt 4: Executive Briefing

“Generate a board-ready incident briefing for the CISO covering: what happened, business impact, containment status, recovery timeline, and HIPAA notification assessment.”

Copilot output: 1-page executive summary ready for the CISO within 3 minutes of incident opening. vs. 4+ hours of manual report writing.

3 Phase 3: Automated Response & Containment

โš”๏ธ Defender XDR: Automated Investigation & Response (AIR)

  • Auto-isolate devices: All 14 compromised devices isolated from the network within 4 minutes of the unified incident creation
  • Disable accounts: 3 compromised accounts disabled in Entra ID, all active sessions revoked
  • Quarantine malware: Ransomware binary quarantined across all endpoints fleet-wide
  • Block IOCs: C2 domains and IP addresses added as custom indicators (block and alert)

๐Ÿ“ก Sentinel: Logic App Playbook. “Ransomware-ContainAndNotify”

This playbook triggers automatically via a Sentinel automation rule when an incident has the tag “ransomware” and severity “high”:

  1. Block C2 on firewalls: Calls Palo Alto API to add blocking rules for all C2 IPs across all branch firewalls (30 seconds)
  2. Restart backup agents: Executes remote PowerShell on affected servers to restart Veeam services
  3. Create P1 incident ticket: Creates a ServiceNow Severity 1 incident with full alert details, affected asset list, and containment status
  4. Notify stakeholders: Sends Teams messages to the SOC channel, emails the CISO and CIO, and pages the on-call incident commander
  5. Trigger evidence preservation: Initiates MDE investigation package collection on all isolated devices for forensic analysis
  6. Update incident: Adds playbook execution log back to the Sentinel incident as a comment for audit trail

4 Phase 4: Recovery & Restoration

  • Backup restoration: 7 operational backup agents restore encrypted servers within 4 hours (vs. 3–6 weeks without protected backups)
  • Credential rotation: All service accounts and admin credentials rotated across affected systems
  • Device rebuild: 5 devices without clean backups rebuilt from gold images and re-onboarded to MDE
  • Monitoring escalation: Sentinel hunting queries scheduled hourly for 72 hours to detect any persistence mechanisms missed

5 Phase 5: Lessons Learned & Hardening

Post-Incident Review (PIR) conducted 72 hours after recovery:

FindingRoot CauseRemediationMicrosoft Solution
VPN without MFALegacy VPN excluded from CA policiesEnforce MFA on all VPN connectionsEntra ID CA policy
9-day dwell timeNo anomaly detection for unusual logon patternsDeploy Sentinel UEBA analytics for user behaviour baselinesSentinel UEBA
Backup agents disabledBackup service accounts had local admin on all serversRemove local admin, use gMSA accounts, deploy tamper protectionMDE Tamper Protection
Saturday timing exploitedReduced weekend SOC coverageDeploy AIR auto-remediation for ransomware patterns 24/7XDR AIR
Manual C2 blockingNo firewall automationDeploy Sentinel playbook for automatic C2 blockingSentinel + Logic App

๐Ÿ“Š Incident Metrics

4 min
Detection to
containment
4 hours
Full recovery
(vs. 6 weeks)
14/8000
Devices affected
(blast radius limited)
3 min
Board briefing
generated by Copilot
$14.8M
Estimated cost
avoided
๐Ÿ“ง

Use Case 2: Business Email Compromise & Identity Attack

Zero Trust: Verify Explicitly NIST: Detect ยท Respond STRIDE: Spoofing ยท Information Disclosure

๐Ÿ“‹ Enterprise Scenario

Organisation: Global consulting firm ยท 20,000 employees ยท 40 countries ยท M365 E5
Attack: CFO’s credentials harvested via a sophisticated phishing site mimicking the Microsoft 365 login page. Attacker creates inbox rules forwarding financial emails to an external address, impersonates the CFO to request a $2.8M wire transfer from the Finance team, and consents to a malicious OAuth app to maintain persistence.
Zero Trust failure: The CFO’s account did not have phishing-resistant MFA (FIDO2), only SMS-based MFA which was bypassed via adversary-in-the-middle proxy.

1 Detection: Multi-Product Signal Correlation

Defender XDR creates a unified incident correlating 5 signals detected within 2 hours:

SignalProductDetection
Risky sign-inEntra ID ProtectionImpossible travel: sign-in from UK (real) and Nigeria (attacker) 30 min apart
Inbox rule creationMDONew rule forwarding emails containing "wire", "payment", "invoice" to external address
OAuth app consentMDA / App GovernanceUnverified app granted Mail.ReadWrite.All via user consent during the compromised session
Impersonation emailMDOEmail from CFO to Finance requesting urgent wire transfer. sent from attacker session
Anomalous mailbox activitySentinelCustom KQL: 400+ emails read in 30 minutes (baseline: 15/hour for this user)

๐Ÿ“ก Sentinel: BEC Detection KQL

// Sentinel Analytics Rule: Suspicious Inbox Rule + Impossible Travel
let riskyUsers = AADSignInEventsBeta
| where Timestamp > ago(2h)
| where RiskLevelDuringSignIn in ("medium", "high")
| distinct AccountObjectId, AccountUpn;
CloudAppEvents
| where Timestamp > ago(2h)
| where ActionType == "New-InboxRule"
| where AccountObjectId in (riskyUsers)
| extend RuleName = tostring(RawEventData.Parameters[0].Value)
| extend ForwardTo = tostring(RawEventData.Parameters[3].Value)
| where ForwardTo !endswith "@contoso.com" // External forwarding
| project Timestamp, AccountUpn, RuleName, ForwardTo, IPAddress

2 AI Investigation with Security Copilot

Copilot Prompt: “What emails did the attacker read and forward from the CFO’s mailbox?”

Result: Copilot queries MDO and returns: 47 emails read containing financial data, 12 emails forwarded to attacker-controlled address, 3 emails sent impersonating the CFO requesting wire transfers totalling $4.1M.

Copilot Prompt: “What OAuth apps were consented during this compromised session? What data did they access?”

Result: 1 app consented: "DocuSign Helper" (unverified publisher) with Mail.ReadWrite.All. App accessed 340 mailboxes over 2 hours using application-level permissions.

Copilot Prompt: “Generate the HIPAA breach assessment for this incident.”

Result: Copilot analyses the data exposure: 47 financial emails containing PII (client names, SSNs in tax documents). Determines HIPAA notification may not apply (financial data, not PHI), but state breach notification laws require assessment within 72 hours. Generates the legal team briefing.

3 Automated Response

  • XDR AIR: Revokes all CFO sessions, forces password reset, removes malicious inbox rules, soft-deletes the impersonation emails from all recipient mailboxes
  • App Governance: Auto-disables the malicious OAuth app, revokes all permission grants
  • Sentinel Playbook “BEC-Containment”: Sends recall notification for the wire transfer request, alerts the Finance department via Teams with verified instructions to halt any pending transfers, creates a Legal hold on the CFO’s mailbox for eDiscovery

4 Lessons Learned & Zero Trust Hardening

FindingRemediationSolution
SMS MFA bypassed by AiTM proxyDeploy phishing-resistant MFA (FIDO2 keys) for all executivesEntra ID + FIDO2
User consented to malicious OAuth appRestrict user consent to verified publishers; require admin approval for all othersEntra ID + App Governance
No monitoring for bulk email readsDeploy Sentinel analytics rule for anomalous mailbox activity volumeSentinel KQL
Wire transfer request not verifiedImplement out-of-band verification process for transfers >$50KOrganisational process
Financial email forwarding to externalMDO transport rule blocking auto-forward to external for Finance departmentMDO mail flow rule
๐Ÿ‘ค

Use Case 3: Insider Data Exfiltration by Departing Employee

Zero Trust: Least Privilege NIST: Detect ยท Respond STRIDE: Information Disclosure ยท Repudiation

๐Ÿ“‹ Enterprise Scenario

Organisation: Semiconductor company ยท 5,000 employees ยท $50M+ IP in R&D designs
Threat: Senior product engineer submits resignation. Over 10 business days, downloads 2,400 design files from SharePoint, copies to personal USB, emails specs to personal Gmail, uploads source code to personal GitHub.
Challenge: Each individual action appears routine. engineers access design files daily. Only cross-product correlation reveals the exfiltration pattern.

1 Detection: Cross-Product Signal Correlation

DayActivityProductDetection
Day 0Resignation submittedPurview IRM (Workday connector)HR trigger auto-enrolls user in “Departing Employee” policy. Adaptive Protection elevates DLP to “strict.”
Day 2Bulk SharePoint downloads (800 files)MDADownload volume 15x above 90-day baseline. Alert generated.
Day 4USB copy of labeled filesMDE + Purview DLPEndpoint DLP blocks USB copy of “Confidential” labeled files. Alert correlated into XDR incident.
Day 6Email to personal Gmail with attachmentsMDO + Purview DLPDLP policy detects technical drawings in email to external personal address. Email quarantined.
Day 8GitHub upload (500 MB)SentinelCustom KQL analytics rule detects large upload to github.com/[personal-account]

๐Ÿ“ก Sentinel: Source Code Upload Detection KQL

// Detect large uploads to personal code repositories
DeviceNetworkEvents
| where Timestamp > ago(24h)
| where RemoteUrl has_any ("github.com","gitlab.com","bitbucket.org")
| where RemoteUrl !has "contoso" // Exclude corporate repos
| summarize TotalBytesSent = sum(SentBytes),
            UploadCount = count() by DeviceName, AccountName, RemoteUrl
| where TotalBytesSent > 100000000 // 100 MB threshold
| extend UploadMB = round(TotalBytesSent / 1048576.0, 1)
| join kind=inner (
    IdentityInfo | where Department == "Engineering"
) on $left.AccountName == $right.AccountName
| project Timestamp=now(), AccountName, Department, 
          RemoteUrl, UploadMB, DeviceName

2 Security Copilot: Investigation Assistance

“Quantify the data exposure: How many of the 2,400 downloaded files had Confidential or Highly Confidential sensitivity labels?”

Result: 847 files labeled Confidential (chip design schematics), 23 labeled Highly Confidential (manufacturing process IP). Estimated IP value at risk: $12M based on R&D investment records.

“Generate the legal evidence package with chain-of-custody documentation for all exfiltration activities.”

Result: Copilot compiles timestamped activity log, DLP policy match records, USB device identifiers, recipient email addresses, and GitHub commit hashes. formatted for legal counsel with chain-of-custody metadata.

3 Response & Legal Action

  1. Adaptive Protection automatically restricts the user’s DLP: USB blocked, external email quarantined, cloud uploads monitored
  2. Purview eDiscovery Premium case created with legal hold on all user data (email, OneDrive, Teams, SharePoint activity)
  3. Sentinel playbook generates evidence package for Legal and sends notification to HR
  4. HR conducts exit interview with documented findings; Legal issues cease-and-desist and takedown request to GitHub

4 Lessons Learned

FindingRemediationSolution
No HR system integrationConnect Workday to Insider Risk Management for automated departure triggersPurview IRM + HR connector
USB not blocked for departing employeesDeploy Adaptive Protection: automatic strict DLP for elevated-risk usersPurview Adaptive Protection
No code repo monitoringDeploy Sentinel KQL rule for large uploads to personal code reposSentinel analytics rule
Sensitivity labels not enforced on all design filesDeploy auto-labeling policy for all files in Engineering SharePoint librariesPurview auto-labeling
๐Ÿ—๏ธ

Use Case 4: Cloud-Native SOC Transformation & ROI

Zero Trust: All Pillars NIST: Govern ยท Identify ยท Detect ยท Respond

๐Ÿ“‹ Enterprise Scenario

Organisation: Insurance company ยท 15,000 employees ยท 3 data centres ยท $4B revenue
Current state: Splunk Enterprise on-premises ($2.8M/year licence + $500K infrastructure) ยท 4 FTEs managing SIEM infrastructure ยท 150 TB annual ingestion ยท 35% analyst turnover ยท 22-minute average MTTT
Goal: Migrate to Microsoft unified security platform, reduce costs, improve SOC effectiveness, and enable Zero Trust architecture

Migration Roadmap

PhaseDurationActivitiesMicrosoft Solutions
Phase 1: FoundationMonth 1–2Deploy Sentinel workspace. Connect M365 data sources (free ingestion). Enable Defender XDR unified incidents. Deploy Security Copilot with 2 SCUs for pilot team.Sentinel, XDR, Copilot
Phase 2: MigrationMonth 3–4Migrate firewall/proxy logs to Sentinel (Basic Logs tier for high-volume). Translate 80 Splunk SPL rules to Sentinel KQL using SIEM migration tool. Deploy 15 Logic App playbooks replacing Phantom SOAR.Sentinel, Logic Apps
Phase 3: OptimisationMonth 5–6Enable UEBA for user behaviour analytics. Deploy custom workbooks replacing Splunk dashboards. Train all analysts on Copilot promptbooks. Expand Copilot to full SOC (4 SCUs).Sentinel UEBA, Copilot
Phase 4: DecommissionMonth 71-month parallel operation. Validate detection parity. Decommission Splunk. Redeploy 4 infrastructure FTEs to threat hunting and detection engineering.All platforms

SOC Effectiveness: Before vs. After

MetricBefore (Splunk + Legacy)After (XDR + Sentinel + Copilot)Improvement
Mean Time to Triage22 minutes8 minutes (Copilot promptbooks)-64%
Mean Time to Respond4.2 hours45 minutes (AIR + playbooks)-82%
False positive rate35%12% (XDR correlation)-66%
MITRE ATT&CK coverage35%75% (Copilot-generated detections)+114%
Annual SIEM cost$3.3M (licence + infra + staff)$1.3M (Sentinel + SCUs)-61%
Infra management FTEs40 (cloud-native)Redeployed to hunting
New analyst ramp time4–6 months6–8 weeks (Copilot assistance)-67%
Data sources integrated45120+ (300+ connectors available)+167%

ROI Summary for Executive Briefing

$2M
Annual cost
savings
82%
Faster incident
response
4 FTEs
Redeployed to
threat hunting
7 months
Full migration
timeline

Continue Learning