Teams Booking Phishing: A Device Code Labeled as a Booking Reference
Huntress's 12 Nov 2026 report: a fake Teams booking page shows a real Microsoft device code as a booking reference, hidden behind an encrypted lure page.

AegisAI Threat Watch · Credential theft
Teams Booking Phishing: A Device Code Labeled as a Booking Reference
Huntress's post of 12 November 2026 describes a fake Teams scheduler that shows a real Microsoft device code as a booking reference, behind an encrypted page that scanners cannot read.
- What
- A fake Teams booking page shows a real Microsoft device code labelled as a booking reference. Entering it at Microsoft's sign-in hands over a session.
- Who it hit
- An employee at a construction company, via an email posing as a senior vice president at another firm.
- Why filters miss it
- The email holds no attacker domain, the lure page is encrypted, and the sign-in page is Microsoft's own.
- Do today
- Block device code flow in Conditional Access where it is not needed, and require MFA on every account.
Reported by Huntress, 12 November 2026.
A fake Teams booking page showed a real Microsoft device code as a booking reference, and typing it into the genuine sign-in page gave the attacker a session.
A device code is a short code that Microsoft's sign-in flow for input-limited devices shows on one screen so that a person can approve the sign-in on another. A typosquatted domain is a deliberately misspelled web address registered to pass for a real company. A session token is the proof of a completed sign-in that an attacker can use in place of a password.
Analysis based on Huntress research published on 12 November 2026, about an incident the company contained in mid-October at a construction-sector company. The post gives no exact incident date. Huntress credits the lure page screenshot and the sign-in log findings to its own security operations center.
Source: Huntress, "Crafty Microsoft Teams Booking Flow Disguises Device Code Phishing", 12 November 2026.
Key judgments
- The email sender was a typosquat of a real company. Huntress says the message impersonated a senior vice president at another construction firm, and that the lookalike domain was registered eight days before the incident.
- The email body held no attacker domain. Its link had been wrapped by the mail platform's link rewriting, then passed through a Google short link and a Google redirect before the lure page. Huntress says this sidestepped URL reputation checks.
- The lure was a scheduling flow, not a login form. A "Microsoft Teams Meeting Scheduler" with the executive's headshot asked for a date, time and duration, then started a real device code request in the background and showed the code as a "Booking Reference".
- The page was built to look empty to scanners. The kit was stored as AES-256-GCM ciphertext with its key and IV on the same page, and decrypted in the browser. A scanner that does not run JavaScript sees an encrypted blob.
- The attacker kept trying. If Microsoft returned an MFA-related error, the kit restarted the device code session and told the victim that a new verification code was being prepared.
Evaluating email security tools? The 2026 Buyer's Guide to Email Security has nine evaluation areas and the questions to ask vendors. Get the guide.
How does a Teams booking page turn a device code into a booking reference?
Huntress describes six steps from the email to the redirect. Each one looks like routine scheduling until the code is typed into Microsoft's page.
- An email from a typosquatted domain, posing as a senior vice president at another construction company, proposes an introductory call about a supply partnership. The body holds a link but not the lure domain.
- The link had been wrapped by the mail platform's link rewriting, then passed through a Google short link (share.google) and a Google redirect. The final page sits on the typosquatted lure domain.
- The page presents a "Microsoft Teams Meeting Scheduler" with the executive's headshot, an "Introductory call" and fields for a date, time and duration.
- On "Continue", the page asks the attacker's backend to start a device code sign-in with Microsoft. A booking confirmation page then shows the returned code as the booking reference, with a session expiration countdown.
- The confirm button copies the code to the clipboard and opens Microsoft's genuine device login page. The victim pastes the code and approves the sign-in, which the attacker started.
- The page then shows "Verification complete" and redirects the victim to what looks like a real Calendly booking page. The attacker holds a session token.
Source: Huntress, 12 November 2026, attack flow and kit analysis.
What a target sees at each step
The table pairs what the employee experiences with what Huntress says is happening. It is AegisAI's arrangement of the Huntress description.
| Step | What the target sees | What is happening |
|---|---|---|
| A call request from a senior executive at a partner firm | A typosquatted domain registered eight days earlier | |
| Link | One click, no unfamiliar domain in the message | A wrapped link that hops through Google services to the lure |
| Scheduler | A Teams-style booking page with a headshot | A small loader that decrypts the real kit in the browser |
| Code | A "Booking Reference" with a countdown | A live Microsoft device code that the attacker requested |
| Sign-in | Microsoft's genuine sign-in page | The victim approves the attacker's session |
| Exit | "Verification complete", then a Calendly page | The attacker holds a token and the visit ends like ordinary scheduling |
How does the encrypted lure page avoid scanners?
The landing page was a small loader. Huntress says the actual kit was stored as AES-256-GCM ciphertext and decrypted in the browser with the WebCrypto API, with the key and IV on the same page. The loader also hooked window.open and two fetch calls to route requests to the attacker's backend.
Encryption did not protect the kit from a person with a browser. It protected it from crawlers and scanners that fetch a page without running its JavaScript: they receive a blob of ciphertext with nothing to match. The decryption key was the weak point only for an analyst who ran the page.
The extra prompt and the retry
Huntress found two behaviours built for failure. If the backend supplied a mirror_url, a pop-up told the victim that the sign-in was successful and that "an additional verification" was needed on the device. Huntress could not see where mirror_url pointed.
If Microsoft returned an MFA-related error, the kit restarted the device code session, told the victim that a new verification code was being prepared, and showed another code. A failed first attempt therefore did not end the lure.
What did the compromise look like in the sign-in logs?
Huntress points to three traces. The value Cmsi:Cmsi in the logs marks a successful device code authentication. The attacker's activity came through a residential proxy that the post places in New Orleans, on what Huntress calls a random home internet address. Automated user agents, python-requests and node, also appeared, often in non-interactive sign-in logs.
The compromised account had no MFA enrolled and was used from a non-compliant, unmanaged device. Huntress also saw the network (ASN) and region change within the same session ID, a sign of the token being reused from elsewhere.
Why the flow can be phished at all
The device authorization grant, standardized as RFC 8628, was written for devices that are input constrained, where typing a password is impractical. The user enters a code on a second device. The RFC's own security section lists remote phishing as a known attack: an attacker emails the target and tells them to enter the user code. It says the code's lifetime should be short enough to limit the usefulness of a code obtained for phishing.
How can an organization stop device code phishing?
Block the flow where it is not needed, and make the account worth little if a code is approved. Huntress and Microsoft's documentation point to controls at five places: the message, the page, the flow, the account and the logs.
| Control | What it sees in this incident | Result |
|---|---|---|
| Sender and domain reputation | A lookalike domain registered eight days earlier, with no history | Can flag it, if age and lookalike are weighed |
| URL reputation on the email link | A wrapped link through Google services, no attacker domain in the body | Huntress says it sidestepped these checks |
| Page scanning that does not run JavaScript | An encrypted blob | Nothing to match |
| Conditional Access block on device code flow | The code is never accepted for ordinary users | Can stop the sign-in |
| MFA on every account | The compromised account had none enrolled | Raises the cost of the attack |
| Sign-in log review | Cmsi:Cmsi, scripted user agents, a new region inside one session | Can reveal the compromise |
| Reading the request in the message | A booking request that leads to a code to type into a sign-in page | Can be flagged when it arrives by email |
AegisAI's reading of the reported chain. The flow and MFA rows follow Huntress and Microsoft Learn.
A booking request, a booking reference and a sign-in on the real Microsoft page. Every step looks like scheduling, until the code is typed.
AegisAI Threat IntelligenceWhat this means for a security team
This section is AegisAI's analysis. Huntress reports one incident at one construction-sector company, but the method does not depend on the sector: any employee who can receive a meeting request can be sent a device code. The lure carries no file and no password form, so there is nothing for a user to recognize as a credential request.
Three questions settle the exposure: whether device code flow is blocked for users who have no headless device, whether every account has MFA enrolled, and whether the team reads sign-in logs for device code sign-ins from unmanaged devices.
Why standard controls miss it
This section is AegisAI's analysis. Each layer sees a harmless piece. The sender domain is new, but nothing in the message is malicious. The link passes through well-known services. The page is ciphertext to a scanner. The sign-in page is Microsoft's own, and the code is real.
How AegisAI protects against it
AegisAI has not observed this campaign. Huntress reports that the lure arrived as an email with a link, so the approach below applies where a booking or meeting request reaches a mailbox.
AegisAI's agents evaluate the message and the destination it points to together, and pull a malicious message out of mailboxes before users see or click it. More on how that works: inbound email security.
When a lure or a compromised session is caught, the response compounds: the message is withheld at classification, copies already delivered are retracted, rewritten links refuse a late visit, the account's outbound and internal mail gets rescanned, and lateral phishing attempts to colleagues are blocked at the same bar rather than treated as a fresh, unrelated message.
Recommendations for defenders
- Block device code flow in Conditional Access for all users, with a report-only trial first and documented exceptions for devices that need it, as Microsoft's guidance recommends.
- Enroll every account in MFA, and prefer phishing-resistant methods. Restrict sign-ins from non-compliant, unmanaged devices.
- Alert on device code sign-ins, the
Cmsi:Cmsivalue, scripted user agents such aspython-requestsandnode, and a change of network or region inside one session. - Train staff to treat any code that a web page asks them to type into a Microsoft sign-in page as an attack, and to confirm unexpected meeting requests with the sender through a known contact.
- After a suspected compromise, revoke sessions and check for rogue MFA methods, unfamiliar devices, new app consents, mailbox rules and outbound mail.
FAQ
What is Microsoft Teams booking device code phishing?
It is a scam in which a fake Teams meeting scheduler starts a real Microsoft device code sign-in and shows the code as a booking reference. When the victim types it into Microsoft's genuine device login page, the attacker's sign-in is approved and the attacker gets a session token. Huntress reported one such incident on 12 November 2026.
How do I protect Microsoft 365 from device code phishing?
Block device code flow with a Conditional Access policy for users who do not need it, trying the policy in report-only mode first. Require MFA on every account, restrict unmanaged devices, and review sign-in logs for device code sign-ins. Microsoft's documentation recommends getting as close as possible to a block on the flow.
How can I tell that a booking reference is really a device code?
A real meeting booking does not ask a person to copy a code and paste it into a Microsoft sign-in page. If a page shows a short code and then opens a Microsoft device login page, stop and confirm the request with the sender through a known contact. Report the page to the security team.
What should I do if I already typed the code?
Tell the security team at once. Huntress advises checking for rogue MFA methods, unrecognized devices, unauthorized app consents, suspicious mailbox rules and suspicious outbound mail. The team should also revoke the account's sessions, because a token can keep working after a password change.
For researchers
What the reporting does not tell us 5 gaps
Actor. Huntress names no group, and does not say whether the kit is sold as a service or used by one actor.
Scale. The post describes one incident at one company. It does not say how many other organizations received the lure.
Dates. Huntress gives "mid-October" for the incident and no exact date. It does not say how long the attacker held the session before containment.
Countdown and mirror_url. The post does not state the length of the countdown, and Huntress could not see where mirror_url pointed.
Mail path. The post says the link sidestepped URL reputation checks. It names no email security product, so which controls the message passed is not reported.
Reported indicators 4 items
Indicators from the Huntress post, defanged. Infrastructure changes quickly: treat these as historical indicators, not a live blocklist.
| Type | Indicator | Role |
|---|---|---|
| Domain | gpimmlgration[.]com | Attacker backend and phishing infrastructure |
| URL | share[.]google/?q=UltJkMsVWI5V9GJM1 | Google short link in the redirect chain |
| Log value | Cmsi:Cmsi | Marks a successful device code authentication |
| User agents | python-requests, node | Automated activity in sign-in logs, often in non-interactive logs |
MITRE ATT&CK mapping 7 techniques
These techniques are AegisAI's mapping of the Huntress description. IDs and names were checked against attack.mitre.org on 15 November 2026.
| ID | Name | Use in this campaign |
|---|---|---|
T1583.001 | Acquire Infrastructure: Domains | A typosquatted domain was registered eight days before the incident. |
T1566.002 | Phishing: Spearphishing Link | An email posing as a senior executive carried a wrapped link to the lure. |
T1608.005 | Stage Capabilities: Link Target | The Teams-style scheduler page was staged on the lure domain. |
T1027.013 | Obfuscated Files or Information: Encrypted/Encoded File | The kit was AES-256-GCM ciphertext, decrypted in the browser. |
T1528 | Steal Application Access Token | The approved device code gave the attacker a token for the account. |
T1078.004 | Valid Accounts: Cloud Accounts | The attacker used the compromised cloud account. |
T1090 | Proxy | The attacker's final sign-in came through a residential proxy. |
References 4 sources
- Huntress (12 November 2026). Crafty Microsoft Teams Booking Flow Disguises Device Code Phishing. huntress.com
- IETF (August 2019). RFC 8628: OAuth 2.0 Device Authorization Grant, section 5.4, Remote Phishing. rfc-editor.org
- Microsoft Learn. Block authentication flows with Conditional Access policy. learn.microsoft.com
- MITRE ATT&CK. Techniques T1583.001, T1566.002, T1608.005, T1027.013, T1528, T1078.004 and T1090. attack.mitre.org
Every email, investigated.
When a booking request with a trusted-looking link reaches a work inbox, see how AegisAI reads the request and where the link leads.
Phishing that reaches the inbox, stopped before anyone clicks.
Book a demoRelated posts
More from the AegisAI blog.

