Microsoft 365 accounts are now usually taken over by getting around multifactor authentication (MFA), not by guessing passwords. Attackers steal the session token after a user completes MFA, trick users into approving sign-ins or malicious apps, then use inbox rules to hide while they redirect payments. The controls that stop this are phishing-resistant MFA enforced through conditional access, blocking device code flow and legacy sign-in, restricting app consent, and watching for the signals described below.
This guide is written for owners, finance leaders and IT managers at Canadian organizations that run on Microsoft 365. It covers the current attack chain, what Microsoft recommends today, what to look for in the logs and what to do in the first hours after a takeover.
Why MFA Alone No Longer Stops Account Takeover
MFA still matters. Microsoft states that, based on its experience, more than 99.9% of common identity attacks such as password spray are stopped by MFA and blocking legacy authentication (Microsoft Learn, 2026). The problem is that the attacks causing real losses are no longer the common ones.
The Canadian Centre for Cyber Security warns that attackers bypass MFA through MFA fatigue, token theft and machine-in-the-middle attacks, and recommends phishing-resistant, FIDO-based methods (Canadian Centre for Cyber Security, 2024). In practice, an account takeover today is a session problem as much as a password problem.
The business impact is usually payment fraud. The Canadian Anti-Fraud Centre (CAFC) reports that spear phishing losses exceeded $68 million (CAD) in 2025 and reached about $57.7 million in the first six months of 2026 alone, often through compromised or spoofed communications used to redirect payments (Canadian Anti-Fraud Centre, 2026). A taken-over mailbox is the most convincing way to run that fraud.
The Attack Chain as of November 2026
Most takeovers combine two or three of the techniques below. Knowing which one was used decides which control would have stopped it.
Adversary-in-the-Middle Phishing and Token Theft
In adversary-in-the-middle (AiTM) phishing, the fake page sits between the user and the real Microsoft sign-in. The user enters a password and approves MFA as normal, and the phishing kit captures the session token that Microsoft issues. Microsoft describes these attacks as intercepting authentication traffic in real time, bypassing MFA that is not phishing-resistant (Microsoft Security Blog, 2026).
Between April 14 and 16, 2026, Microsoft observed one AiTM campaign, built around fake "code of conduct" notices, targeting more than 35,000 users across more than 13,000 organizations in 26 countries (Microsoft Security Blog, 2026).
Microsoft's incident guidance defines token theft as replaying tokens issued to a user even when that user has satisfied MFA, which grants the attacker access because the authentication requirements were already met (Microsoft Learn, 2026).
Device Code Phishing
Device code sign-in exists for TVs, conference room devices and command-line tools that cannot show a browser. Attackers start a genuine device code request, then persuade the victim to enter the code on Microsoft's real sign-in page. The attacker then receives valid access and refresh tokens (Microsoft Security Blog, 2025). Because the victim signs in on the real Microsoft site, URL checks and most user training do not help.
MFA Fatigue
MFA fatigue, also called push bombing, is when an attacker who already has the password sends repeated push prompts until the user accepts one (Canadian Centre for Cyber Security, 2024). Microsoft now enables number matching for all Microsoft Authenticator push notifications, which means the user must type a number shown on the sign-in screen (Microsoft Learn, 2026). That blunts blind approvals, but it is not phishing-resistant: a user can still be persuaded to complete a sign-in the attacker started.
OAuth Consent Phishing
Consent phishing tricks a user into granting permissions to a malicious cloud application, which then reaches mail and files without needing the password (Microsoft Learn, 2026). By default, Microsoft Entra ID lets all users consent to applications for permissions that do not need administrator approval, including access to their own mailbox (Microsoft Learn, 2026). A password reset does not remove an app the user has approved.
Inbox Rules and the Move to Payment Fraud
Once inside, attackers hide. Microsoft notes that attackers create rules that delete messages, move them to little-used folders such as RSS, mark them as read or forward them externally, sometimes keyed to words such as "invoice" or "phish" (Microsoft Learn, 2026). The CAFC describes the same pattern in Canadian cases: fraudsters create rules to forward copies of incoming email, study payment schedules and contacts, then ask accounts payable to pay a supplier invoice into a new bank account (Canadian Anti-Fraud Centre, 2026). Our business email compromise guide covers the payment side in depth.
The Controls That Stop It: Microsoft's Current Guidance
The controls below map directly to the techniques above. Licensing differs between them, so check which Microsoft licence each one needs before planning.
Require Phishing-Resistant MFA with Conditional Access
Phishing-resistant MFA is the control that defeats AiTM, because the credential is bound to the real sign-in site. In Microsoft Entra, the built-in "Phishing-resistant MFA strength" accepts Windows Hello for Business or platform credentials, FIDO2 security keys (including passkeys) and multifactor certificate-based authentication (Microsoft Learn, 2026). You apply it through a conditional access policy, which requires a Microsoft Entra ID P1 licence (Microsoft Learn, 2026).
Microsoft's own guidance after the April 2026 campaign was to enable passwordless methods such as Windows Hello, FIDO keys or Microsoft Authenticator, and to scope conditional access so privileged accounts must use phishing-resistant MFA (Microsoft Security Blog, 2026). A practical order: administrators first, then finance, executives and anyone who approves payments, then everyone else. For why text codes fall short, see is SMS MFA good enough?
Use Security Defaults If You Have No Conditional Access
If your tenant has no Entra ID P1 licences, turn on security defaults. It requires all users to register for MFA, requires MFA for administrators, blocks legacy authentication protocols and blocks device code flow (Microsoft Learn, 2026). Microsoft notes that legacy protocols such as IMAP, POP3 and older SMTP clients do not support MFA, so an attacker can use them to bypass it (Microsoft Learn, 2026). Security defaults is a floor, not a ceiling: it cannot require phishing-resistant methods for specific groups.
Block Device Code Flow and Legacy Authentication
Microsoft recommends that organizations get as close as possible to a full block on device code flow and allow it only for documented, secured uses (Microsoft Learn, 2026). Starting July 1, 2026, new Microsoft Entra tenants block device code flow as part of security defaults (Microsoft Learn, 2026). Existing tenants that use conditional access should add the block themselves, starting in report-only mode to find any conference room devices that still depend on it.
Turn on Token Protection Where Your Devices Support It
Token protection is a conditional access session control that accepts only sign-in tokens bound to the user's registered device, so a stolen token cannot be replayed from another machine (Microsoft Learn, 2026). As of November 14, 2026, Microsoft lists it as generally available for native applications on Windows, iOS, iPadOS and macOS, with enforcement for Exchange Online, SharePoint Online and Teams. Apple devices must be MDM-managed with Microsoft's Enterprise SSO plug-in, and browser-based support is in preview and limited to selected web apps that access Azure Resource Manager, so web access to mail and files is not yet covered (Microsoft Learn, 2026). Microsoft advises piloting in report-only mode first (Microsoft Learn, 2026). Treat it as a second layer behind phishing-resistant MFA, not a replacement.
Restrict User Consent to Applications
Microsoft recommends allowing user consent only for applications from verified publishers (Microsoft Learn, 2026). Pair that with an admin consent workflow so staff can request an app instead of approving it themselves, and review existing app permissions once a quarter.
Strengthen the Email Layer
Microsoft's recommendations also include Safe Links and Safe Attachments in Defender for Office 365, zero-hour auto purge to pull delivered phishing out of mailboxes, network protection in Defender for Endpoint and automatic attack disruption in Defender XDR (Microsoft Security Blog, 2026). These reduce how many lures reach users; they do not replace identity controls.
Detection Signals to Watch
Prevention fails sometimes. Microsoft's token theft playbook lists the signals that should trigger an investigation (Microsoft Learn, 2026):
- Risky sign-in detectionssuch as anomalous token, unfamiliar sign-in properties and unfamiliar non-interactive sign-ins (some require Entra ID P2).
- New MFA methods or passwordless credentialsadded to an account, especially shortly after a sign-in from a new location.
- Mailbox forwarding rules added or edited, and inbox rules that delete, hide or mark messages as read.
- Newly registered devicesor credential changes in the audit log.
- Unusual or mass file downloadsfrom OneDrive or SharePoint.
- Clicks on suspicious URLsthat may lead to an AiTM kit.
Add two business signals: a supplier or client asking why you sent unusual messages, and any request to change banking details. Those are often the first sign a mailbox has been taken over.
These signals only help if someone reviews them quickly, around the clock. Sign-in and audit logs should feed a monitored system, not sit unread in an admin portal.
Recovery Checklist: If an Account Is Taken Over
Speed matters, because the attacker may be preparing a payment request. Work through these steps in order, and see what to do if Microsoft 365 is hacked for a shorter version.
- Block sign-in and revoke sessions. In the Microsoft Entra admin portal, disable the account and select Revoke sessions, which stops new tokens being issued (Microsoft Learn, 2026). Access tokens issued by Entra ID last one hour by default, and sessions issued by individual applications may need to be ended in those applications (Microsoft Learn, 2026).
- Reset the password after sessions are revoked, not before.
- Remove attacker persistence. Delete suspicious inbox rules and forwarding, remove MFA methods and devices the user did not add, and revoke consent for unfamiliar applications.
- Contain the payment risk. Warn finance staff and the suppliers and clients the mailbox corresponds with. If money has moved, call your bank immediately to request a recall, then report it. The CAFC credits timely reporting with the January 2026 freeze of a $1.7 million transfer linked to spear phishing (Canadian Anti-Fraud Centre, 2026).
- Scope the incident. Check sign-in logs for other accounts accessed from the same IP addresses, and search for phishing sent from the compromised mailbox to colleagues.
- Preserve evidence. Export audit and sign-in logs before retention limits remove them; they matter for insurers and any privacy assessment.
- Assess privacy obligations. Under PIPEDA, organizations must report breaches of security safeguards that pose a real risk of significant harm to the Privacy Commissioner of Canada, notify affected individuals as soon as feasible, and keep records of all breaches for two years (Office of the Privacy Commissioner of Canada, 2025). Provincial privacy laws can add their own obligations; our privacy breach response guide explains them.
- Report. Report the incident to the Canadian Centre for Cyber Security and, where fraud or crime is involved, to local police (Canadian Centre for Cyber Security, 2026).
- Close the gap. Identify which technique was used and apply the matching control from the section above before you call the incident closed.
If you are in the middle of an incident now, go to our emergency page.
What to Do Now
A short plan for the next 30 days:
- Confirm whether you have Entra ID P1 licences. If not, turn on security defaults today.
- Create conditional access policies requiring phishing-resistant MFA for administrators, then for finance and executives.
- Block device code flow and legacy authentication, using report-only mode first.
- Set user consent to verified publishers only and turn on the admin consent workflow.
- Review every mailbox for external forwarding and unusual inbox rules.
- Pilot token protection on managed Windows devices.
- Write down the recovery steps above, name who performs each one, and rehearse them once.
- Require a call-back to a known number before any change to supplier or payroll banking details.
How Cyberphore Can Help
Cyberphore hardens Microsoft 365 tenants, including conditional access, phishing-resistant MFA for administrators and finance, and legacy protocol removal, then monitors sign-ins, mailbox rules and app consents 24/7/365, with critical alerts triaged and response beginning within 15 minutes. Learn more about our Microsoft 365 and email security service.



