Mon–Sat 10:00–18:00 London · UK
Remote & on-site ☎ 0207 096 0936
← Guides
Security

Microsoft 365 Account Hacked? What to Do in the First Hour

Reset the password and revoke every session straight away, because a reset on its own can leave the attacker signed in. Then remove the MFA methods, inbox rules and app permissions they added, find out what was sent, and warn anyone who might act on a fake payment request.

A customer rings to ask about the strange email you sent them this morning. Or a member of staff can’t sign in, because the password they have used for years has suddenly stopped working. Either way, assume someone else is inside the mailbox.

Work in order. Get the attacker out, then remove what they left behind, find out what they sent, and warn the people who received it. An administrator can do most of this within an hour from the Microsoft 365 and Entra admin centres.

Confirm the signs

Microsoft lists these symptoms:

  • people receiving messages the owner didn’t send, often with a link or a request to pay;
  • the mailbox suddenly unable to send;
  • inbox rules nobody remembers creating, or mail going missing;
  • unfamiliar messages in Sent Items or Deleted Items;
  • unexplained password changes or lockouts;
  • an altered email signature, or new forwarding to an outside address.

One of these is enough to act on.

1. Reset the password and revoke every session

In the Microsoft 365 admin centre, go to Users > Active users, select the person and choose Reset password. Then open the user again and, on the Account tab, select Sign out of all sessions. The Entra admin centre has the same control as Revoke sessions, on the user’s overview page under Entra ID > Users.

Both steps matter. A reset stops the old password working, but it doesn’t reliably end sessions that already exist: Microsoft’s token documentation lists kinds of session that survive one. That gap is worst when a fake sign-in page captured the session itself, which puts the attacker past MFA already. Revoking sessions invalidates the account’s refresh tokens, so every app must sign in again. Even that can take up to an hour, the life of an access token, and Outlook on the web may hang on until the page is refreshed.

Two details catch people out:

  • Give the new password by phone or in person, never by email. The attacker may still be reading it.
  • Delete any app passwords. Microsoft notes a password reset doesn’t revoke them.

2. Check the MFA methods

Attackers often register their own phone or authenticator app so they can pass MFA later. In the Entra admin centre, go to Entra ID > Users, select the person and open Authentication methods. Delete anything the owner doesn’t recognise. Require re-register MFA clears phone numbers, Authenticator registrations and software tokens in one go, but it doesn’t cover every method type, so check the list afterwards. If self-service password reset is switched on, a rogue phone number here is a route to resetting the password you’ve just changed.

Have the person re-register while you’re on the phone, so you know the new method is theirs.

3. Hunt down inbox rules and forwarding

This is where most of the damage hides. The usual pattern is a rule that watches for words like “invoice”, “payment” or “bank”, then deletes the message, marks it as read, or moves it to a folder nobody opens, such as RSS Subscriptions, Archive or Junk Email. The owner never sees the supplier’s genuine reply, and the attacker carries on the conversation.

Outlook on the web shows rules under Settings > Mail > Rules, but some never appear there. An administrator can list every rule, hidden ones included, by running Get-InboxRule with the -IncludeHidden switch in Exchange Online PowerShell. Check the mailbox’s own forwarding setting too. Our explainer on why Microsoft 365 blocks email auto-forwarding covers where forwarding hides and how the tenant-wide block works.

Before deleting anything, write down what each rule matched and where it sent mail. That tells you what may already have left the business.

4. Review app consents and admin roles

A quieter route skips the password entirely. The attacker persuades the user to approve a third-party app, which then holds its own permission to read or send mail. That consent stays granted after a password reset until someone removes it. In the Entra admin centre, go to Entra ID > Enterprise apps > All applications, open anything unfamiliar and select Permissions, where the User consent tab shows what individuals approved. The portal can’t revoke user consents (that takes PowerShell or Microsoft Graph), but you can stop the app at once: under Properties, set Enabled for users to sign-in? to No.

Check, too, that the account hasn’t gained admin roles.

5. Find out what was sent

Look through Sent Items and Deleted Items for messages the owner didn’t write. Attackers often delete what they sent, so tidy folders prove little. Message trace in the Exchange admin centre, under Mail flow > Message trace, works from the service’s delivery records rather than the mailbox, so it still shows those messages. The last 10 days come back instantly, and up to 90 days are available as a downloadable report. Export the recipient list for step 8.

6. Get the mailbox sending again

If the account sent spam, Microsoft may have blocked it from sending. Its messages bounce with error 5.1.8, though it still receives mail. Once the account is secure, open the Microsoft Defender portal, go to Email & collaboration > Review > Restricted entities, select the mailbox and choose Unblock. Microsoft says the restriction usually lifts within an hour and within 24 hours at most. Unblock too early and the attacker can get it blocked again.

7. Read the sign-in and audit logs

The Entra sign-in logs, under Entra ID > Monitoring & health > Sign-in logs, show when the account signed in, from which IP address and location, and whether it succeeded. Check the non-interactive tab too. Look early, because retention depends on licence and is short: seven days on the free tier of Entra ID that comes with Business Basic and Business Standard, and 30 days with the Entra ID P1 included in Business Premium.

The unified audit log, searched from Audit in the Microsoft Purview portal, keeps records for 180 days on most business plans, and a year for Exchange, SharePoint and Entra activity by users with E5 licences. It records sign-ins and the creation of inbox rules, so it can show when the attacker got in and when the rules appeared. New events can take an hour or more to appear.

8. Tell the people who received it

Phone the people on the recipient list from step 5, starting with anyone due to pay you or be paid by you. Tell them plainly that any change of bank details from your address is fake, and that you’ll confirm payment details by phone. An email from the same mailbox isn’t enough, because it looks exactly like the fraud did. Our guide to invoice fraud and business email compromise covers the payment checks that stop the next attempt. If personal data may have been exposed, the reporting duties, including the ICO’s 72-hour deadline, are in our data breach checklist.

Then close the doors

  • MFA for everyone, with no standing exceptions.
  • Block legacy authentication. Older mail protocols can’t do MFA, and Microsoft found more than 99% of password-spray attacks use them.
  • Security defaults or Conditional Access. Security defaults are free on every plan, make everyone register for MFA and block legacy authentication. Conditional Access needs Entra ID P1, which Business Premium includes, and replaces them with rules you can tailor. You can’t run both at once.

Getting these settled is part of a proper Microsoft 365 setup. If you’re mid-incident and would rather not work through it alone, cyber security help in the first hour makes the rest of the week easier.

Frequently asked questions

Should we disable the account while we investigate, or let the person keep working?

Microsoft's guidance prefers disabling the account until the investigation is finished, and if the person can manage without email for a few hours, that is the cleaner option. Blocking sign-in from the Microsoft 365 admin centre can take up to 24 hours to apply, so reset the password and revoke sessions as well rather than relying on the block. If they must keep working, finish the MFA, rule and app checks before they sign back in, and have them do it on a device you trust.

Who gets told when Microsoft blocks a mailbox for sending spam?

A default alert policy called "User restricted from sending email" notifies administrators when it happens. Out of the box it goes to the Global Administrators, which in a small business may be one person, or an account nobody reads. Find it in the Microsoft Defender portal under Email & collaboration > Policies & rules > Alert policy and add a recipient who checks their inbox. Before unblocking a sender, Microsoft also suggests running its Blocked Sender History diagnostic, which shows past blocks and whether the account has MFA.

How do we know whether other accounts have been hit too?

Take the IP addresses from the hacked account's suspicious sign-ins and search the sign-in logs for other users signing in from them, which is what Microsoft's own investigation guidance suggests. Then check message trace for anything the account sent to colleagues, because an email from a workmate is easy to trust, and ask anyone who clicked a link in one to say so. Any account that turns up gets the same treatment, in the same order.

Our accounts sync from a server in the office. Does that change the steps?

Yes. If accounts sync from an on-premises Active Directory, reset the password there rather than in Microsoft 365, and Microsoft recommends resetting it twice to guard against pass-the-hash attacks. If sign-in is federated through your own servers, the change has to happen on-premises as well. Revoking sessions, checking MFA methods and removing rules still happens in the cloud admin centres, exactly as described here.

Related services

Free · no obligation

Want a hand with any of this?

Tell us what you're trying to sort out and we'll come back with a clear, no-obligation plan and price.