Back to Intelligence

Microsoft Entra Connect Health Exploited to Steal Tenant Data: Detection and Hardening Guide for Identity Defenders

SA
Security Arsenal Team
November 22, 2026
10 min read

Microsoft has disclosed that attackers exploited a security flaw in its Entra Connect Health service to gain unauthorized access to administrator email addresses and tenant configuration information. This is not a theoretical exposure — Microsoft confirmed unauthorized access occurred, which means affected tenant data is already in adversary hands.

For defenders, the immediate concern is not just what was taken, but what it enables. Administrator email addresses and tenant configuration data are reconnaissance gold. They fuel highly targeted phishing against privileged accounts, password-spray campaigns against known-valid identities, and social engineering calibrated to your exact hybrid identity architecture. If your organization runs Entra Connect (formerly Azure AD Connect) with the Health agent deployed — and most hybrid environments do — you should assume elevated targeting risk for your privileged identity population in the coming weeks.

This post breaks down the exposure, gives you concrete hunt queries for your SIEM and endpoints, and walks through the hardening steps I'd prioritize on an IR retainer call today.

Technical Analysis

What Was Exposed

Based on Microsoft's disclosure, the compromised data set includes:

  • Administrator email addresses — a ready-made target list of privileged identities in affected tenants
  • Tenant configuration information — details about how the tenant and its hybrid identity synchronization are configured

Why Entra Connect Health Matters in Your Architecture

Entra Connect Health is the monitoring agent that ships alongside Entra Connect synchronization. It runs on-premises (as the AzureADConnectHealth* services) and continuously reports sync health, AD FS health, and domain controller telemetry up to the Microsoft cloud. Because it sits at the seam between your on-prem Active Directory and your Entra ID tenant, it has visibility into — and trust relationships with — the most sensitive tier of your identity infrastructure.

A vulnerability in this service is strategically valuable to attackers for three reasons:

  1. Ubiquity: Nearly every hybrid-joined enterprise runs the Health agent.
  2. Privilege adjacency: The sync infrastructure it monitors operates with powerful credentials (the MSOL_* / Sync_* connector accounts) that can read and write directory objects.
  3. Low defender visibility: Many SOCs treat the Health agent as plumbing and don't baseline its behavior.

Exploitation Status

Microsoft has confirmed actual unauthorized access — this is a breach disclosure, not a pre-emptive patch advisory. No public proof-of-concept or CVE identifier has been published in the source reporting at time of writing. Defenders should treat the flaw as confirmed exploited and prioritize the follow-on-attack risk (targeted phishing, credential attacks against exposed admin identities) rather than waiting for technical exploitation details to leak.

The Likely Follow-On Attack Chain

Given the stolen data (admin emails + tenant configuration), the realistic adversary playbook is:

  1. Spear-phish exposed admin identities with tenant-specific lures (fake Entra sync alerts, credential-expiry notices referencing your actual configuration).
  2. Attempt MFA fatigue or token theft against those accounts.
  3. On success, pivot to persistence — adding federated domains, modifying directory sync settings, or registering malicious service principals.

Every detection below is built around that chain.

Detection & Response

Sigma Rules

The highest-fidelity detections here live in your Entra ID logs, not on endpoints. The first rule targets anomalous behavior of the on-prem directory synchronization account — the crown jewel adjacent to the compromised service. The second catches identity-infrastructure persistence changes that follow a successful admin compromise. The third flags bulk directory reads consistent with tenant enumeration.

YAML
---
title: Anomalous Sign-In by Entra Connect Sync Account
id: 3f7a2b91-4c5d-4e8a-9b1c-6d2e8f0a1b2c
status: experimental
description: Detects failed or anomalous interactive sign-in attempts against the Entra Connect directory synchronization account (Sync_* or MSOL_*), which should only authenticate from the on-prem sync server for synchronization operations.
references:
  - https://attack.mitre.org/techniques/T1078/
  - https://www.bleepingcomputer.com/news/security/microsoft-entra-connect-health-flaw-exploited-to-steal-tenant-data/
author: Security Arsenal
date: 2026/02/10
tags:
  - attack.initial_access
  - attack.t1078
  - attack.persistence
logsource:
  product: azure
  service: signinlogs
detection:
  selection_account:
    UserPrincipalName|startswith:
      - 'Sync_'
      - 'MSOL_'
  selection_anomaly:
    ResultType|contains:
      - '50053' # Account locked
      - '50055' # Password expired
      - '50057' # Account disabled
      - '50126' # Invalid credentials
      - '70008' # Expired/revoked refresh token
  condition: selection_account and selection_anomaly
falsepositives:
  - Password rotation of the sync account performed by administrators during maintenance
level: high
---
title: Federation or Domain Authentication Configuration Change in Entra Tenant
id: 8c4d5e6f-7a8b-4c9d-0e1f-2a3b4c5d6e7f
status: experimental
description: Detects changes to domain federation or authentication settings in an Entra ID tenant, a common persistence technique following compromise of administrative credentials such as those exposed in the Entra Connect Health breach.
references:
  - https://attack.mitre.org/techniques/T1484/002/
  - https://www.bleepingcomputer.com/news/security/microsoft-entra-connect-health-flaw-exploited-to-steal-tenant-data/
author: Security Arsenal
date: 2026/02/10
tags:
  - attack.persistence
  - attack.t1484.002
  - attack.defense_evasion
logsource:
  product: azure
  service: auditlogs
detection:
  selection:
    OperationName|contains:
      - 'Set domain authentication'
      - 'Set federation settings on domain'
      - 'Update domain'
      - 'Add verified domain'
      - 'Set Company DirSync Enabled Status'
  condition: selection
falsepositives:
  - Legitimate domain onboarding or migration projects; validate against change management records
level: critical
---
title: Bulk Directory Object Read via Service Principal
id: 1a2b3c4d-5e6f-4a5b-8c9d-0e1f2a3b4c5d
status: experimental
description: Detects service principals performing directory read operations consistent with tenant enumeration (user, group, and role harvesting) that may follow compromise of tenant configuration data.
references:
  - https://attack.mitre.org/techniques/T1087/004/
  - https://www.bleepingcomputer.com/news/security/microsoft-entra-connect-health-flaw-exploited-to-steal-tenant-data/
author: Security Arsenal
date: 2026/02/10
tags:
  - attack.discovery
  - attack.t1087.004
logsource:
  product: azure
  service: auditlogs
detection:
  selection:
    OperationName|contains:
      - 'Add service principal'
      - 'Add app role assignment to service principal'
      - 'Consent to application'
    TargetResources|contains:
      - 'Directory.Read.All'
      - 'RoleManagement.ReadWrite.Directory'
      - 'User.Read.All'
  condition: selection
falsepositives:
  - Legitimate enterprise application onboarding; verify against approved app registrations
level: high

KQL — Microsoft Sentinel / Defender Hunting

This query joins sign-in anomalies for sync accounts with identity-infrastructure configuration changes, giving you a single pane for the post-compromise behaviors most likely to follow this breach. Run it over at least 30 days, and baseline the sync account's expected source IPs first.

KQL — Microsoft Sentinel / Defender
let Lookback = 30d;
// Part 1: Anomalous sign-ins against directory sync accounts
let SyncAccountAnomalies =
    SigninLogs
    | where TimeGenerated > ago(Lookback)
    | where UserPrincipalName startswith "Sync_" or UserPrincipalName startswith "MSOL_"
    | where ResultType != 0
       or IPAddress !in ((SigninLogs
           | where TimeGenerated between (ago(90d) .. ago(30d))
           | where UserPrincipalName startswith "Sync_" or UserPrincipalName startswith "MSOL_"
           | where ResultType == 0
           | summarize make_set(IPAddress)))
    | summarize FailedAttempts=count(), DistinctIPs=dcount(IPAddress),
      IPs=make_set(IPAddress), Locations=make_set(Location),
      FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated)
      by UserPrincipalName, AppDisplayName;
// Part 2: Identity infrastructure changes in the tenant
let IdentityConfigChanges =
    AuditLogs
    | where TimeGenerated > ago(Lookback)
    | where OperationName has_any (
        "Set domain authentication",
        "Set federation settings on domain",
        "Update domain",
        "Add verified domain",
        "Set Company DirSync Enabled Status",
        "Add service principal",
        "Consent to application")
    | extend InitiatedBy = tostring(InitiatedBy.user.userPrincipalName)
    | extend TargetResource = tostring(TargetResources[0].displayName)
    | project TimeGenerated, OperationName, InitiatedBy, TargetResource,
      IPAddress = tostring(InitiatedBy.user.ipAddress), Result, CorrelationId;
union SyncAccountAnomalies, IdentityConfigChanges
| order by TimeGenerated desc

If you ingest Entra ID sign-in data via CommonSecurityLog (e.g., through a CEF connector), adapt the sync-account portion to filter DestinationUserName startswith "Sync_" on those events instead.

Velociraptor VQL — Endpoint Hunt for Entra Connect Health Agents

If Microsoft releases an updated Health agent, your first question will be: where are these agents, and what version are they running? This artifact inventories the Entra Connect Health agent services across your fleet and pulls their binary versions so you can scope remediation without waiting on your CMDB.

VQL — Velociraptor
-- Inventory Entra Connect Health agent processes, services, and installed versions
LET processes = SELECT Pid, Name, Exe, Username
FROM pslist()
WHERE Name =~ 'AzureADConnectHealth'
   OR Exe =~ '(?i)Azure AD Connect Health'

LET binaries = SELECT FullPath, Mtime, Size
FROM glob(globs='C:/Program Files/Azure Ad Connect Health */**/*.exe')

SELECT 'process' AS ArtifactType, Name AS Detail, Exe AS Path,
       Username AS Context, '' AS Version
FROM processes
UNION ALL
SELECT 'binary' AS ArtifactType, basename(path=FullPath) AS Detail,
       FullPath AS Path, str(string=Mtime) AS Context,
       version_info(path=FullPath).ProductVersion AS Version
FROM binaries

Remediation & Verification Script

Use this on your Entra Connect / Health agent servers and from an admin workstation to verify agent inventory, review privileged role assignments for anomalies, and force session revocation for privileged accounts as a precautionary measure following the data exposure.

PowerShell
# ============================================================
# Entra Connect Health breach response - verification & hardening
# Run Section 1 on sync servers; Sections 2-4 from an admin workstation
# Requires: Microsoft.Graph PowerShell SDK, Global Reader+ (PIM-elevated)
# ============================================================

# --- Section 1: Inventory Entra Connect Health agents and versions (on sync server)
Get-Service -Name "AzureADConnectHealth*" -ErrorAction SilentlyContinue |
    Select-Object Name, Status, StartType | Format-Table -AutoSize
Get-ItemProperty "HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*",
                 "HKLM:\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*" `
    -ErrorAction SilentlyContinue |
    Where-Object { $_.DisplayName -like "*Azure AD Connect Health*" } |
    Select-Object DisplayName, DisplayVersion, InstallDate | Format-List

# --- Section 2: Review privileged role assignments for unexpected accounts
Connect-MgGraph -Scopes "RoleManagement.Read.Directory","Directory.Read.All","AuditLog.Read.All" -NoWelcome
Get-MgRoleManagementDirectoryRoleAssignment -All | ForEach-Object {
    $principal = Get-MgDirectoryObject -DirectoryObjectId $_.PrincipalId -ErrorAction SilentlyContinue
    [PSCustomObject]@{
        RoleDefinitionId = $_.RoleDefinitionId
        PrincipalId      = $_.PrincipalId
        PrincipalName    = $principal.AdditionalProperties.userPrincipalName
        PrincipalType    = $principal.AdditionalProperties.'@odata.type'
    }
} | Export-Csv -Path ".\PrivilegedRoleAssignments_$(Get-Date -Format 'yyyyMMdd').csv" -NoTypeInformation
# Compare against your last known-good export; investigate any delta immediately.

# --- Section 3: Identify directory sync accounts and check for unauthorized credential changes
Get-MgUser -Filter "startswith(userPrincipalName,'Sync_') or startswith(userPrincipalName,'MSOL_')" -All |
    Select-Object UserPrincipalName, AccountEnabled, LastPasswordChangeDateTime, CreatedDateTime |
    Format-Table -AutoSize

# --- Section 4: Precautionary session revocation for privileged accounts exposed in the breach
# Feed in your admin list (or the list Microsoft provided via your tenant notification)
$AdminAccounts = Get-Content ".\exposed_admin_upns.txt"   # one UPN per line
foreach ($upn in $AdminAccounts) {
    try {
        $user = Get-MgUser -UserId $upn -ErrorAction Stop
        Revoke-MgUserSignInSession -UserId $user.Id
        Write-Host "[+] Revoked sessions for $upn" -ForegroundColor Green
    } catch {
        Write-Warning "[-] Failed to process $upn : $_"
    }
}

# --- Section 5: Audit recent directory sync configuration changes
Get-MgAuditLogDirectoryAudit -Filter "activityDisplayName eq 'Set Company DirSync Enabled Status'" -Top 20 |
    Select-Object ActivityDateTime, ActivityDisplayName, Result, InitiatedBy |
    Format-List

Remediation

  1. Apply Microsoft's guidance immediately. Monitor the Microsoft Security Response Center and your Microsoft 365 Message Center / Entra admin center for tenant-specific notifications. Microsoft has stated it is notifying affected customers directly — confirm whether your tenant is on that list before assuming you're clear.

  2. Update the Entra Connect Health agent on all sync servers as soon as Microsoft publishes a fixed build. Verify deployment with the script in Section 1 above — don't trust your CMDB for this.

  3. Reset credentials for the directory synchronization accounts (Sync_* / MSOL_*) in your tenant. These accounts sit directly adjacent to the compromised service. Re-run the Entra Connect wizard or use Set-ADSyncAADCompanyFeature workflows per Microsoft's documented rotation procedure.

  4. Enforce phishing-resistant MFA on all administrator accounts now. The stolen admin email list will be phished. FIDO2/passkeys or certificate-based auth for every privileged role — no exceptions for break-glass accounts (store those credentials offline instead).

  5. Revoke active sessions and refresh tokens for privileged identities as a precaution (Section 4 of the script). Token theft is the most common follow-on to admin-targeted phishing.

  6. Audit identity-infrastructure configuration: review federated domain settings, directory sync configuration, service principal consent grants, and privileged role assignments for unauthorized changes in the last 30 days. The Sigma rules and KQL above operationalize this.

  7. Alert your service desk and admins to expect targeted phishing referencing Entra Connect, sync errors, or credential expiry — lures that leverage the stolen tenant configuration data will look unusually credible.

  8. Tune conditional access: restrict admin portal access to compliant devices and trusted network locations, and enable Continuous Access Evaluation to shorten the window on any stolen tokens.

Executive Takeaway

The data stolen here is reconnaissance material, not the endgame. The breach's true cost will be measured in the phishing campaigns and identity attacks it enables over the next 90 days. Organizations that treat this as a prompt to harden privileged identity — phishing-resistant MFA, session revocation, sync-account rotation, and active hunting for federation tampering — will blunt the follow-on wave. Those that file it as "Microsoft's problem" will find out it wasn't.

Related Resources

Security Arsenal Healthcare Cybersecurity AlertMonitor Platform Book a SOC Assessment healthcare Intel Hub

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.