Trapdex Subscribe

CampaignsConsent and approval

EvilTokens: phishing kit gets Microsoft 365 users to enter a device code on the real Microsoft sign-in page, handing over their session

  • Status unknownWe can't confirm whether this has been taken down. Treat it as possibly live. Status checked .
  • Moderate confidence
  • Seen to

Do this now

  1. Block the device code flow with a Conditional Access authentication-flows policy, with an allow-list only for devices that need it
  2. Hunt Entra sign-in logs for authenticationProtocol deviceCode over the last 90 days; revoke sessions and review new device registrations for any hit
  3. Tell staff that no document, bid or voicemail ever needs them to type a code into a Microsoft page

What happened

EvilTokens, a phishing-as-a-service kit advertised on Telegram from 16 Feb 2026, sent tailored lures (construction bids, DocuSign, Microsoft Forms, shared files) that ended with the victim typing an attacker-supplied code into the genuine Microsoft device login page. That signs the attacker's session in as the victim. Huntress counted more than 340 affected organizations in the US, Canada, Australia, New Zealand and Germany, with tokens replayed from Railway.com infrastructure.

  • Email
  • Microsoft 365
  • Microsoft Entra ID

Attack chain

  1. Lure

    Business-themed email (bid request, DocuSign, voicemail, shared file) routed through redirects to a code prompt

    First seen

  2. Artifact

    The user enters a device code at the real Microsoft device login page, approving the attacker's sign-in

    The victim approved access on a genuine sign-in or consent page. Nothing was installed on the device.

  3. Actor

    No source we cite attributes this activity to an actor.

  4. Infrastructure

    • Redirect chains through compromised sites and short-lived hosting (workers.dev, Vercel, Amplify, App Engine)
    • Token replay from Railway.com hosting (a legitimate platform abused by customers of the kit)

    Last seen

Timeline

  1. Proofpoint reports
  2. First seen
  3. Huntress reports
  4. Last seen
  5. Cloud Security Alliance reports
  6. Status checked: Status unknown

None yet.

Indicators

No public indicators.

Confidence: Moderate

Strong evidence, not yet confirmed by our own analysis. See the confidence scale for what each level requires. The sources below are the evidence.

Capped at Moderate: there is no follow-on payload with a published hash that we can analyze.

Sources

Notes

  • EvilTokens is a kit sold to many customers, not an actor, so actor is null. The 340+ count is from Huntress. CSA (25 Mar 2026) relays it and is the readable copy, because the Huntress page could not be opened for a check on 16 Nov 2026.
  • Same technique, separate actors (not tied to EvilTokens by their sources): TA2723 and UNK_AcademicFlare (Proofpoint, 18 Dec 2025); Storm-2372 (Microsoft, 13 Feb 2025).
  • Capped at Moderate. There is no follow-on payload with a published hash, and Huntress publishes no kit-specific domains, so indicators is empty. Railway, workers.dev, Vercel and Amplify are shared platforms and are not listed.
  • last_seen is Huntress's 23 Mar 2026 update, which says Huntress was still blocking attempts from Railway infrastructure that day and ties the campaign to the EvilTokens kit.
  • Status checked 16 Nov 2026. The Huntress page could not be opened for a check that day. No source we cite reports an end to the activity, and no hashes are recorded. Status stays unknown.

Corrections

No corrections.

See the corrections policy. Spotted an error? Tell us.