What happened
Replaying the Storm-2372 device code campaign
- 1
Lure: a conversation and a meeting invitation
From Aug 2024; dates of single lures not statedThe actor builds rapport with the target on WhatsApp, Signal or Teams, posing as a prominent person, then sends an invitation to an online event or meeting that asks the user to enter an actor-generated device code.
- 2
Tokens: a real sign-in page, the actor's code
From Aug 2024; dates of single sign-ins not statedThe target enters the actor's device code on a legitimate Microsoft sign-in page, approving MFA if prompted, and the actor receives the access and refresh tokens; a device code expires 15 minutes after creation, per Volexity.
- 3
Mail: Microsoft Graph with the actor's tokens
From Aug 2024; dates not statedThe actor uses Microsoft Graph with the tokens to search mail for keywords such as credentials, admin and password and to exfiltrate the matching messages, through proxies matched to the target's region.
- 4
Device registration: a registered device and a Primary Refresh Token
Observed in Feb 2025From February 2025 the actor uses the Authentication Broker client ID to get a refresh token, registers an actor-controlled device in Entra ID and obtains a Primary Refresh Token to collect mail.
Accounts of organizations in government, NGO, IT, defense, telecom, health, education and energy sectors were opened with tokens that targets handed over through a real sign-in page, and mail was searched and exfiltrated; from February 2025 the actor also registered its own device in Entra ID to obtain a Primary Refresh Token and collect mail. The number of victims, the end of the activity and what was taken are not stated.
Where Microsoft Entra controls fit
A Conditional Access policy that blocks the device code flow would have refused the sign-in the target completed, so no token would have reached the actor through that flow, and the device registration that started from a device code sign-in would have been refused with it. The policy does not act on a mail search made with a token that already exists. Turning off user registration of devices, and requiring a compliant device for mail, would have bounded the device registration step, and token protection could have refused mail requests made with tokens that are not bound to a device, where it applies. The chat lure and the user's choice sit outside every cloud control. Check whether your tenant still allows the device code flow for users who do not need it.
Could this have been prevented?
In hindsight, not as a whole. 2 controls would have refused a step outright, and 4 controls would have limited one. 2 steps have no cloud-provider control. The counts below are control assessments, not attack stages: several can address the same step. The same replay with the controls in place before the incident shows where each phase would have been refused or bounded.
- 2
- Prevented
- 4
- Contained
- 2
- Not covered
How it could have been prevented
The same replay with the controls in place. Green marks where a step is refused, blue where it is bounded, and red where no control reaches it. Open a row to read the control, why it acts, and its caveat.
Replaying the Storm-2372 device code campaign
1 · Lure: a conversation and a meeting invitation
1 Not covered2 · Tokens: a real sign-in page, the actor's code
1 Prevented1 Not covered3 on other clouds3 · Mail: Microsoft Graph with the actor's tokens
2 Contained4 · Device registration: a registered device and a Primary Refresh Token
1 Prevented2 ContainedKey facts
- Microsoft says Storm-2372 activity started in August 2024; the targets were government, NGO, IT services and technology, defense, telecommunications, health, higher education, and energy and oil and gas organizations in Europe, North America, Africa and the Middle East; it names no organization and gives no count of victims
- The actor posed as a prominent person relevant to the target on WhatsApp, Signal or Teams, then sent invitations to online events or meetings, including fake Teams meetings, that asked the user to enter an actor-generated device code
- The target entered the code on a legitimate Microsoft sign-in page and the actor received the access and refresh tokens; Volexity, describing campaigns under its own labels, reports that a device code expires 15 minutes after creation and that the target approved MFA if prompted
- The actor used Microsoft Graph to search mail for keywords such as credentials, admin and password and exfiltrated matching mail, and used proxies matched to each target's region to hide sign-ins; Volexity reports downloads of OneDrive files for its actors
- From February 2025 the actor used the Microsoft Authentication Broker client ID to obtain a refresh token, registered an actor-controlled device in Entra ID and obtained a Primary Refresh Token; Microsoft's update of February 14, 2025 says the actor began using that client ID in device code flows within the previous 24 hours
- Microsoft first assessed with moderate confidence that Storm-2372 aligns with Russian interests; its update of July 31, 2026 assesses it as an initial access operations sub-cluster of Midnight Blizzard, with no confidence level stated
Sources
Every fact in this breakdown comes from the public reporting below. The replay is a model of that reporting, not a reconstruction of the environment it describes.
- 1.Storm-2372 conducts device code phishing campaignMicrosoft Threat Intelligence · Feb 13, 2025 · Vendor reportThe actor, attribution, start of activity, targets, lures, the device code flow, device registration and the Primary Refresh Token, mitigations
- 2.Multiple Russian threat actors targeting Microsoft Device Code AuthenticationVolexity · Feb 13, 2025 · ResearchThe device code flow steps, the code lifetime, MFA approved by the victim, the Conditional Access block, the timing of the January 2025 campaigns
The estate in the replay is illustrative and drawn at the level of attack stages; the victim's real environment is not public. Verdicts describe what a control does when it is in place before the incident. They are not an assessment of the victim's security program.