Fake authentication attacks steal Microsoft 365 accounts at scale

Fake authentication attacks steal Microsoft 365 accounts at scale

Rating
—/5 0 ratings
Follow us on Google News
Technology8 min read

A purported helpdesk call followed by an SMS about updating a passkey or multi-factor authentication can lead to a Microsoft 365 account takeover even when an employee does not disclose their password. In campaigns observed since May 2026, attackers used passkey, MFA and single sign-on pretexts to persuade victims to authorise a session or device under their control.

In brief: passkeys have not been “cracked”. The risk arises when someone is persuaded to confirm a sign-in request they did not initiate. In some cases, the page where the victim enters the code is the genuine Microsoft page. The issue is who generated the code and for which application or device the access token will be issued.

Microsoft describes this mechanism in its analysis of passkey-themed social engineering and cloud identity compromise. Separately, Huntress researchers documented a device-code phishing campaign targeting more than 340 Microsoft 365 organisations. There is no public basis for treating the two campaigns as the same operation, although the technical mechanism being exploited is similar.

Microsoft 365 attacks where the official page can be part of the trap

In conventional phishing, an attacker seeks to steal a password through a fake sign-in page. In recent attacks, the focus shifts from the password to authorisation: the victim is instructed to approve a code, an MFA request or an OAuth flow initiated elsewhere.

The device authorization grant, commonly known as the device-code flow, is legitimate. It was designed for devices without a keyboard or a full browser, such as a television, video-conferencing device or command-line tool. The device displays a code, and the user enters it in a browser on another device to sign in.

The IETF’s RFC 8628 explains the separation between the device initiating the request and the device on which the user completes authentication. This separation is useful in normal use, but it can be abused: an attacker starts the flow on their own infrastructure, then sends the victim the code and persuades them to enter it on the official page. Once authorised, the tokens are delivered to the client that initiated the flow — in other words, the attacker.

What appears to be happening What may technically be happening What to check
“IT is updating your passkey or MFA.” The victim is given a pretext to start or approve an unknown authorisation. Did the user open the request, or did it follow an unsolicited call or SMS?
A genuine Microsoft page is opened. The page may correctly validate the identity, but the code may be associated with a client initiated by the attacker. Who owns the device, application or session for which the code is requested?
The user approves MFA as usual. MFA may confirm the victim’s authentication in a fraudulent OAuth flow, without having been “cracked”. Is there an MFA notification or sign-in the user did not initiate?
The sign-in appears complete. The attacker may use the token to access resources permitted to that account. Have unusual sessions, MFA methods, applications or file activity appeared?

“Protection” is the pretext; the access token is the target

An authentic passkey is a cryptographic credential based on a key pair and associated with the website or application for which it was created. This mechanism is specifically designed to resist phishing: the private key is not sent to the server. A basic explanation is available from the FIDO Alliance.

This makes an important distinction necessary: the fact that a criminal uses the word “passkey” in a call or message does not mean they have found a vulnerability in passkey technology. In the campaign described by Microsoft, the term was a credible pretext to rush the victim into an adversary-in-the-middle attack or into authorising a device code.

The distinction also matters for organisations that have begun adopting passwordless authentication. Passkeys remain more resistant to phishing than passwords, but no authentication mechanism can fully eliminate the risk of an employee being manipulated into approving an operation they do not understand.

What an attacker can do after authorisation

Following unusual authentications, Microsoft reported reconnaissance activity through Microsoft Graph, access to and downloads from SharePoint and OneDrive, and email collection through REST APIs. Authentication methods added by attackers were also observed.

These observations do not mean that every compromised account automatically had all its data exfiltrated. There is a difference between gaining access, persistence, resource discovery, data collection and confirmed exfiltration. For the victim, however, a valid token can be highly valuable: it provides access within that account’s permissions, without the attacker necessarily needing the original password again.

An important signal may emerge after the purported “security update”, rather than in the original message. Users should treat any new MFA method, unknown device, approval request they did not initiate or session from an unusual context as an incident. Administrators should correlate these signals with access to mailboxes, Microsoft Graph, SharePoint and OneDrive.

Why the phone call matters for organisations in Romania

The public sources consulted for this article do not confirm that the same Microsoft campaign using a passkey pretext has compromised victims in Romania. The technique is nevertheless locally relevant to any organisation using Microsoft 365, particularly because it often begins outside corporate email: through a call, SMS or conversation on a personal phone.

According to a DNSC statement reported by AGERPRES, 42% of incidents reported in Romania in the first half of 2026 were phone scams, compared with 14% phishing and 13% smishing. The figures do not measure this Microsoft campaign and should not be read as evidence of local victims. They do, however, show why the purported caller’s identity should be verified separately, through an internal number or a known channel.

What to do if you entered a code or approved a suspicious request

Speed matters, but the response should not be improvised. For a work account, the employee should immediately inform the IT team or security officer through a known organisational channel, not the phone number used in the call.

  • Stop interacting with the caller and do not enter further codes or approve MFA notifications.
  • Record the time of the call, the phone number, the message received and what was approved; these details may help the investigation.
  • Change the password only through the official channel designated by the organisation and after reporting the incident; changing the password alone may not invalidate an already issued token.
  • Check the authentication methods and devices associated with the account. Any unknown method should be reported, not hastily removed before the IT team can understand the incident.
  • Request revocation of active sessions and tokens, as well as a review of inbox rules, applications with permissions and unusual file activity.

For personal Microsoft accounts, the same principle applies: use only the official account-security page, change the password if justified, revoke sessions and review sign-in methods. If that account is also used for work, the organisation must be notified immediately.

What administrators can do without unnecessarily disrupting work

Microsoft allows the device-code flow to be controlled through Conditional Access and recommends blocking it where there is no operational need. Its documentation on authentication flows in Conditional Access warns, however, that some Teams equipment, CLI tools or older applications may depend on this method.

For that reason, a global shutdown without an inventory can cause disruption. A prudent approach is to begin in report-only mode, identify the applications and groups legitimately using the flow, then introduce narrow, documented exceptions. At the same time, the organisation should monitor device-code authorisations, authentication-method registrations, unusual OAuth consents and large-scale access to cloud data.

The internal message to employees should be very specific: the helpdesk does not ask by phone or SMS for sign-in codes, unsolicited MFA approvals or completion of a “passkey activation” flow initiated by the caller. Independently verifying the call is safer than judging the voice, urgency or appearance of a Microsoft page.

The scale of the problem must be read correctly

In 2026, Huntress documented a separate operation called EvilTokens, which used device-code phishing against more than 340 Microsoft 365 organisations in several countries. The report confirms that abuse of legitimate authorisation flows can be industrialised and can avoid conventional password theft.

The figure is not, however, a measure of the passkey-themed campaign described by Microsoft in September 2026. The two sets of information should remain distinct until public evidence emerges of the same infrastructure, operators or victims. Automatically linking them to other trends, including stolen access to paid AI tools, would require the same type of evidence; the sources consulted contain no sufficient public confirmation for such an attribution.

The practical conclusion is less dramatic but more useful: the correct address of a sign-in page is not the only safety criterion. Before entering a code, approving MFA or registering a new method, the key question is whether the user initiated that process themselves and knows exactly which device or application they are authorising.

Frequently asked questions

Were passkeys cracked in these attacks?

There is no evidence that passkeys were cracked. In the campaigns described, the term “passkey” was used as a social-engineering pretext to obtain authorisation for a session or OAuth flow initiated by the attacker.

Is it safe to enter the code on the official Microsoft page?

Not automatically. In a device-code phishing attack, the Microsoft page may be genuine, but the code may have been generated by a client controlled by the attacker. What matters is who initiated the process and what the code authorises.

Does MFA protect against device-code phishing?

MFA remains an important protection, but it can be approved by the victim in a fraudulent flow. The attack does not crack MFA; it exploits the user’s confirmation of a request initiated by the attacker.

For more information about DNSC, see New legislation for routers: What you need to know, which provides complementary details on this subject.

Can an organisation block the device-code flow?

Yes. Microsoft Entra allows this flow to be controlled through Conditional Access. Before blocking it, legitimate uses must be inventoried, as some Teams devices, CLI tools or older applications may depend on it.

Sources consulted

Etichete: Multi-factor authenticationDNSCMicrosoft 365OAuthPasskeysPhishingcybersecurity
Share this article
COMUNITATEA INFOBUZ

Discuția cititorilor

Ai o opinie sau o întrebare despre acest articol? Spune-ne ce crezi.

0 comentarii

Fii primul care deschide discuția.

Lasă un comentariu

Email address este vizibilă doar administratorilor și nu va fi publicată. Câmpurile marcate cu * sunt obligatorii.

Te rugăm să păstrezi un limbaj civilizat și să discuți la subiect. Comentariile spam sunt filtrate automat.