Our Thinking / Field Notes / The Security Watch

EVOLVING SECURITY ADVISORY / MICROSOFT ENTRA CONNECT HEALTH API VULNERABILITY AND TENANT METADATA EXPOSURE

Microsoft Entra Connect Health API Flaw: What Was Exposed and What to Check

Microsoft has confirmed unauthorised read-only access to tenant metadata through a now-fixed Entra Connect Health API flaw. Here is what was exposed, what was not, what remains unknown and which checks are worth doing.

What matters now

Current position: Fixed server-side on 19 November 2026. Residual risk is targeted phishing or social engineering using exposed tenant metadata and administrator contact addresses.

Affected versions: Microsoft-hosted Microsoft Entra Connect Health API service. Introduced 23 July 2026 and fixed 19 November 2026. No customer-installed software version is identified as affected.

Observed exploitation: Microsoft has confirmed unauthorised read-only access to tenant-level metadata for notified organisations. Attacker identity, number of affected tenants and exploitation period are not disclosed. No unauthorised configuration changes were identified by Microsoft.

Recommended next action: No customer patch required. Check Azure Service Health for a notice, review Entra audit and sign-in logs, brief administrators and technical contacts on targeted phishing, and enforce phishing-resistant MFA for privileged accounts.

Immediate guidance

No customer patch is needed. Microsoft fixed the Entra Connect Health API flaw on its own servers on 19 November 2026 and says unauthorised parties obtained read-only access to tenant-level metadata for affected organisations, including administrator and technical contact email addresses. Microsoft says passwords, password hashes, authentication tokens and valid cryptographic keys were not exposed. No CVE has been identified. The practical risk is targeted phishing of the people named in your tenant, so check Azure Service Health for a notice, review Entra audit and sign-in logs, brief privileged staff and make sure administrator accounts use phishing-resistant MFA.

Where things stand on 22 November 2026

Microsoft has told affected customers, through Azure Service Health, that unauthorised read-only access to tenant-level metadata had occurred via a vulnerability in the Microsoft Entra Connect Health API. Microsoft fixed the flaw on the service side on 19 November 2026 and finished its investigation on 20 November 2026. There is nothing for customers to install.

Microsoft has not publicly disclosed the technical root cause or provided further details to the press. The cited reporting has not identified a CVE or CVSS score for this incident. The account below is based on Microsoft's customer notification as quoted by reporters, and may be updated as more information becomes available.

For decision-makers, the short version is this: it is a disclosure of contact and configuration details, not a confirmed theft of credentials. That still deserves attention, because administrator email addresses combined with accurate tenant details make a very good script for a convincing phishing email.

What Entra Connect Health does

Many UK organisations still run Active Directory on their own servers and link it to Microsoft Entra ID so staff have one identity for both worlds. Microsoft Entra Connect Health is the monitoring layer for that link. A small agent runs on your on-premises identity servers and reports health data to a cloud service. You then see alerts, performance data, usage analytics and synchronisation errors in the Microsoft Entra admin center.

It covers three kinds of server: Microsoft Entra Connect (the sync server), Active Directory Federation Services (AD FS) and Active Directory Domain Services (domain controllers). Microsoft's documentation says it needs Microsoft Entra ID P1 licensing, and that agents update themselves automatically. If you have never opened the Connect Health page, you may still be affected, because the metadata concerned sits at tenant level and a managed service provider may have set it up for you.

Timeline, and what nobody outside Microsoft knows yet

The dates come from Microsoft's customer notification as reported by BleepingComputer and SecurityOnline. The flaw was introduced on 23 July 2026 and Microsoft detected it on 9 November 2026, about 109 days later. A server-side fix followed on 19 November, and Microsoft completed its investigation and began notifying affected organisations through Azure Service Health on 20 November. The issue became widely public on 21 November.

Three gaps matter. Microsoft has not said who exploited the flaw. It has not said how many tenants were affected; press reports mention more than a dozen administrators saying they received notices, and some managed service providers received several, but that is anecdote rather than a count. And Microsoft has not said when exploitation began.

That last point is easy to misread. The 23 July date marks when the bug was introduced, not when anyone first abused it, so the 109 days is the detection gap and not an attack duration. A notice from Microsoft is the best indicator that your tenant was accessed, but without a published count nobody can say how widely the flaw was used.

Some reports use the word stolen, and some say attackers could potentially have modified monitoring configurations. Microsoft's own finding is read-only access with no unauthorised configuration changes identified. We have used Microsoft's wording and treat modification as a theoretical capability rather than something that happened.

What was exposed, and what Microsoft says was not

Microsoft's notification lists the exposed information as metadata about your tenant and its Connect Health setup. That is a different category from stolen credentials, and the distinction is the heart of this incident. Metadata tells someone who you are and how your identity estate is arranged. Credentials let them act as you.

Microsoft also states that its review of available telemetry found no unauthorised configuration changes.

  • Exposed: email addresses of tenant technical contacts and administrator accounts.
  • Exposed: tenant ID, display name, initial domain name, country code and regional information.
  • Exposed: Microsoft Entra licence tier, plus directory synchronisation status, including whether sync is enabled and the last sync timestamp.
  • Exposed: Connect Health onboarding status.
  • Exposed: an expired certificate once used to register Connect Health agents with a decommissioned Microsoft internal service. Microsoft says it is no longer in use but the API still returned it.
  • Not exposed, per Microsoft: passwords, password hashes, authentication tokens, valid cryptographic keys, on-premises server or Active Directory topology, and Azure subscription information.

Why harmless-looking metadata is still useful to an attacker

Phishing succeeds when a message feels expected. A generic 'your account is locked' email is easy to ignore. A message that addresses a named administrator, uses the correct tenant display name and initial domain, and refers to directory synchronisation or Connect Health, is far harder to dismiss. The licence tier even hints at which security features the target may or may not have.

This is our engineering judgement rather than a reported attack. Microsoft has not said the data has been used for phishing, and we are not aware of any public evidence linking it to a specific campaign. But the exposed fields are exactly what a social-engineering attempt aimed at privileged accounts would want, and the people it identifies are the ones with the most power in your tenant.

Expect lures such as fake Microsoft security notices about this very incident, requests to 're-register' a Connect Health agent, or urgent calls to 'verify' administrator access. Verify any notice by going to the Azure portal directly, never through a link in an email.

Checks to do now

None of these needs a maintenance window. Most take minutes. If you are an MSP, work through them for each customer tenant separately, as notices arrive per tenant.

  1. Look for the notice. In the Azure portal, open Azure Service Health and check your health advisories and security advisories for a message about the Entra Connect Health API. Go there directly rather than through any link in an email. A notice means Microsoft confirmed access affecting your tenant. No notice is reassuring but not a formal all-clear, since Microsoft has not published a list of affected tenants.
  2. Review who is named. Open Microsoft Entra Connect Health, then check the notification settings (under Sync errors, Notification settings) and the role assignments under Role based access control (IAM). Remove leavers and unneeded Owner, Contributor and Reader assignments, and make sure the contact addresses are real, monitored mailboxes. Microsoft's documentation says notifications are on by default, Hybrid Identity Administrators have the Owner role, and Global Administrators retain full access even if they do not appear in the role-assignment list.
  3. Review Entra audit and sign-in logs. Look at the administrator accounts and technical contacts for unfamiliar locations, unexpected MFA prompts, new authentication methods, password reset attempts, new app consents or role changes. Treat this as a check for follow-on activity. Microsoft has not said whether the API calls themselves appear in customer logs, so a clean result does not prove your tenant was untouched.
  4. Brief the people who are exposed. Tell Global Administrators, technical contacts, the help desk and anyone who handles identity requests that convincing, accurate-looking emails and calls may arrive. Ask them to report rather than act on anything that mentions Connect Health, sync problems or urgent administrator verification.
  5. Confirm administrator accounts use phishing-resistant MFA. That means FIDO2 security keys, Windows Hello for Business or certificate-based authentication, not SMS or app approval prompts that a convincing fake page can capture. Conditional Access authentication strengths can enforce it for privileged roles. Introduce any new policy in report-only mode first and keep an excluded, protected break-glass account, so a policy mistake does not lock you out.

Running Microsoft's metadata inspection script

Microsoft published read-only PowerShell scripts through aka.ms/ConnectHealth/scripts, which redirects to the AzureAD/EntraIDGovernance repository on GitHub. They send read-only requests to Azure Resource Manager and print JSON. They make no configuration changes.

Understand what the output means before you rely on it. The scripts show the current values returned for your tenant, so they reveal which categories of metadata could have been exposed. They do not show what an attacker actually accessed or when. They are useful for understanding your exposure and for explaining it to others, not for forensics.

You will need Windows PowerShell 5.1 or PowerShell 7 or later, version 2.17.0 or later of the Az.Accounts module, and read access to Microsoft Entra Connect Health in the target tenant. It is worth confirming you can open the Connect Health pages without an access denied error first. Download the scripts from Microsoft's repository to a folder you control, and read them before running them, as you would with any script that signs in to your tenant.

  1. Install or update the Az.Accounts module for your own user account. This only affects your PowerShell profile.
    powershell
    Install-Module -Name Az.Accounts -MinimumVersion '2.17.0' -Repository PSGallery -Scope CurrentUser -ErrorAction Stop
  2. Sign in to the tenant. Replace YOUR_TENANT_ID_OR_DOMAIN with your tenant ID or primary domain. When asked for a subscription, press Enter to accept the default. Microsoft's folder also includes SelectConnectHealthTenant.ps1, which prompts for a tenant ID and signs in for you.
    powershell
    Connect-AzAccount -Tenant 'YOUR_TENANT_ID_OR_DOMAIN'
  3. Run GetTenant.ps1 from the folder containing the downloaded scripts. It returns the tenant-level Connect Health configuration as JSON.
    powershell
    .\GetTenant.ps1
  4. Run GetTenantPksCertificateDetails.ps1. It decodes the stored certificate metadata in memory and does not write certificate files. This is the script relevant to the expired agent-registration certificate.
    powershell
    .\GetTenantPksCertificateDetails.ps1
  5. Read the output for the categories Microsoft listed, such as tenant details, sync status and the expired certificate. Seeing them confirms that this kind of data exists for your tenant. It does not tell you whether it was read. Do not paste output into tickets or chats that others can see, as it may contain your administrator addresses and tenant identifiers.

No patch needed: immediate checks versus routine good practice

The flaw was in Microsoft's cloud API, not in software running on your servers. Microsoft fixed it centrally on 19 November 2026, so there is no agent version, hotfix or configuration change that closes it for you. Connect Health agents also update themselves automatically, and Microsoft's documentation says automatic upgrade cannot be switched off. Treat any message insisting on an urgent patch, installer or agent reinstallation as a likely phishing attempt.

Keep this separate from routine maintenance of Microsoft Entra Connect on your own servers. Keeping the sync server on a supported version is sensible hygiene, but it is unrelated to the Microsoft-hosted API fix. Do not reset credentials, rotate certificates or re-register agents solely on the basis of this incident. Follow any additional tenant-specific guidance from Microsoft.

Separating the two groups below avoids turning a metadata disclosure into a month-long programme. The first group is worth doing this week. The second is good practice that this incident happens to highlight.

  • Immediate: check Azure Service Health for a Connect Health notice in every tenant you manage.
  • Immediate: review Entra audit and sign-in logs for the named administrators and technical contacts, then brief them on targeted phishing.
  • Immediate: confirm privileged accounts use phishing-resistant MFA, with break-glass accounts excluded and protected separately.
  • Immediate: if you are an MSP, check each customer tenant and tell customers whose notices went to your shared mailboxes.
  • Routine: keep administrator and technical contact addresses limited, current and monitored, ideally role-based and separate from everyday mailboxes.
  • Routine: review Connect Health role assignments and notification recipients regularly, and remove monitoring for retired servers.
  • Routine: make sure Azure Service Health alerts reach a mailbox or ticket queue that someone reads, since this notice arrived by that route.
  • Routine: rehearse how staff verify unusual requests from 'Microsoft' by going to the portal directly instead of using the message's link.

The wider lesson on cloud monitoring services

Monitoring and management APIs tend to be treated as low-risk, because they carry health data rather than documents or passwords. This incident is a reminder that they often hold the map of who administers what. A flaw that went undetected for about 109 days in a service most organisations configured once and forgot is a fair argument for treating telemetry services as part of your attack surface.

Microsoft did notify customers directly and fixed the issue within ten days of detection, which is the right behaviour. The limits are also worth noting. No affected-tenant figure, no attacker attribution, no exploitation window and little technical explanation have been given, which makes it hard for customers to judge their own risk. That is a judgement about the available information, not an accusation. We will update this note if Microsoft publishes more, assigns a CVE or revises its statement.

Update history

We’ll keep this page updated as new information becomes available. Check back for the latest developments.

  1. 22 November 2026: Initial update. Microsoft has fixed the service-side flaw (19 November), completed its investigation (20 November) and notified affected customers via Azure Service Health. It reports unauthorised read-only access to tenant metadata, with no exposure of passwords, hashes, tokens, valid keys or on-premises topology, and no unauthorised configuration changes identified. No CVE, attacker identity, affected-tenant count or exploitation window has been published. Update source

Technical terms explained
AD FS
Active Directory Federation Services, a Windows Server role that lets users sign in to cloud and partner applications with their on-premises Active Directory identity.
Azure Service Health
The part of the Azure portal where Microsoft posts service issues, planned maintenance and health or security advisories that affect your tenant or subscriptions.
CVE
A Common Vulnerabilities and Exposures identifier, the standard reference number used to track a specific publicly disclosed security flaw.
metadata
Descriptive information about your tenant and its configuration, such as contact addresses, domains and sync status, as opposed to passwords or the content of your data.
Microsoft Entra Connect Health
A Microsoft cloud monitoring service for on-premises identity servers. Agents on those servers report health, performance and sync data to the Entra admin center.
phishing-resistant MFA
Multi-factor authentication that cannot be tricked into approving a fake sign-in page, such as FIDO2 security keys, Windows Hello for Business or certificate-based authentication.
social-engineering
Manipulating people into handing over access or information, usually by impersonating someone trustworthy, rather than by attacking software directly.
tenant
Your organisation's dedicated instance of Microsoft Entra ID and Microsoft cloud services, identified by a tenant ID and an initial domain name.

Sources and further reading

KEEP THE CONVERSATION GOING

Want help checking your hybrid identity setup?

If you manage Entra ID alongside on-premises Active Directory and want a second pair of eyes on administrator accounts, MFA enforcement or tenant notifications, you can get in touch.

Contact Rismor