Password reset still left mailbox reach intact once a Primary Refresh Token sat in a kit admin panel. BlackApex, DarkApex, SPY, and XXX101 phishing-as-a-service kits lured workforce users with SharePoint, Zoom, OneDrive, and similar themes, coached a Microsoft device-code paste, and let victims finish sign-in on the genuine Microsoft device-login page, MFA included. Attacker backends redeemed that completed code for Microsoft 365 access and refresh tokens, and could capture a Primary Refresh Token used against mailboxes. According to LevelBlue SpiderLabs on 30 October 2026, researchers observed more than 70 kit instances sharing the same backend and admin panel. Phishing-proof device-bound authentication prevents this Microsoft device-code phishing path by removing brokered flows that hand redeemable tokens to a party who never held the enrolled device.
What LevelBlue found across the BlackApex panel cluster
LevelBlue SpiderLabs described a family of panels under names including BlackApex, DarkApex, SPY, and XXX101 rather than a single named-company filing. Delivery rode trusted services first, including Google share links and PDFs that pointed people at Microsoft Forms, before the device-code lure appeared. Kit front ends themed as SharePoint, Zoom, OneDrive, Adobe Acrobat Sign, and Excel copied a Microsoft device code to the clipboard and opened microsoft.com/devicelogin, or ended at login.microsoftonline.com/common/oauth2/deviceauth.
The user pasted the code into Microsoft’s real device-login flow. Sign-in UI could show approval for “Microsoft Authentication Broker” on another device. Authentication, including whatever MFA the tenant required on that genuine page, completed under the victim’s hands. LevelBlue SpiderLabs reported that by then the attacker’s server had redeemed the code and held a live access token and refresh token for the victim’s Microsoft 365 account. The same reporting states the kit also captured a Primary Refresh Token displayed in its administrative panel and used to access the victim’s mailbox.
LevelBlue did not name victim organizations and did not publish victim or record counts. The reporting did not attribute the activity to any specific threat actor. Through 2026, adjacent PhaaS offerings such as EvilTokens, Kali365, and Tycoon2FA also weaponized Microsoft device-code sign-in; LevelBlue states the BlackApex/SPY family does not match those publicly documented kits even though the underlying technique class overlaps. The same brokered-login pattern showed up earlier in Midnight Blizzard device-code phishing against Microsoft 365, where tokens issued after a completed real sign-in again outlived the interactive MFA moment. For the full attack-chain write-up on how these panels cashed in the paste, read the companion on legacymfa.sucks: BlackApex SPY kits redeem M365 device codes.
Why finishing MFA on the real page still lost the mailbox
Workforce users never typed a password into a reverse-proxy clone of the Microsoft form. They completed a real brokered login the kit already controlled through the device code. LevelBlue SpiderLabs put the failure plainly: the victim completed MFA themselves, so the control meant to stop the attacker was satisfied by the wrong person. MFA proved the human on microsoft.com. It did not prove that the party redeeming the code should hold cloud mail.
| Stage | What operators obtained | When it happened relative to sign-in |
|---|---|---|
| Initial access | Access and refresh tokens via code redemption | After victim finished genuine device-login, MFA included |
| Persistence | Primary Refresh Token shown in kit admin panel | After redemption; PRT-backed reach beyond one session |
| Data access | Mailbox reach via redeemed OAuth and PRT material | After tokens were live; no second login required |
Initial access is the only stage where redesigning authentication still changes the outcome. The kit never needed to steal a typed OTP off a fake page. It needed a human to finish Microsoft’s device-login ceremony while the backend polled and exchanged the code. Password and MFA on the real page both ran for the victim. Tokens still issued to whoever held the completed device code.
Persistence followed the Primary Refresh Token. PRT-backed access extended SSO-style reach beyond a single interactive session and, per intake notes on this family, remained usable after a password reset. Helpdesks that treat “reset the password” as full containment are wrong once refresh-class material still authorizes the client the kit controls. Data access was mailbox API use with those redeemed tokens. No second phishable login was required to read mail once access, refresh, and PRT material were live.
Session lifetime tuning only forces faster reuse. It does not undo a redemption that already succeeded, and it does nothing useful if long-lived refresh or PRT-class material still authorizes the attacker's client.
Removing device-code redemption before tokens exist
Phishing-proof, device-bound authentication cannot prevent device-code phishing. The private key never leaves hardware. Signatures bind to the legitimate origin and to an already-enrolled device. They cannot refuse a device-code approval, because the origin is genuine and the tenant still allows the grant. Operators who only open microsoft.com/devicelogin and wait still receive material they can redeem if the device-code flow remains available for standard workforce accounts.
That is prevention before a session exists for the operator, not detection after a PRT appears in a panel. The Prevention, Not Detection model is the point: name a block or limit on device-code sign-in, and close the brokered redemption path. Conditional Access policies that block device-code grant for ordinary users are what refuse the flow the kits productized.
| Control | Against coached device-code completion | Against PRT or mailbox use after redemption |
|---|---|---|
| Password + MFA on real device-login | Victim still finishes sign-in for the kit | Does not revoke issued tokens |
| Passkeys at login only | Hardens the ceremony origin; flow may still exist | Does not erase tokens already issued |
| Device-bound phishing-proof auth | Cannot prevent device-code phishing; does not close the grant | Does not undo tokens already held |
| Token revoke + CAE | N/A after prevention | Required cleanup if material already leaked |
Passkeys and FIDO2-style WebAuthn place a key pair on the device, keep the private key in hardware, and produce an origin-bound signature at login. That is origin-bound authentication against spoofed origins when the ceremony itself is the battleground. A generic enterprise identity provider that issues only hardware-bound signatures shows the login hardening cleanly. Passkeys alone are not documented here as blocking completion on Microsoft’s genuine device-login page when device-code grant remains enabled, and sources do not claim that equivalence. origin-bound login still leaves enrollment, recovery, and brokered grant paths as separate design choices.
LevelBlue's BlackApex write-up did not show helpdesk re-enrollment or recovery abuse for this family. The published path is a coached completion of the real device-login page, then token redemption. Closing that path means blocking the grant, not tightening enrollment after the fact.
Once redemption has already succeeded, login cryptography does not reach backward into mailbox APIs. Explicit revoke of access tokens, refresh tokens, and PRT-class material, plus continuous access evaluation, is containment. Malware that steals cookies from a machine that already holds a legitimate session is a harder residual problem and is not the cheap remote path these panels sold.
Key takeaways for defenders
- Disable or tightly Conditional-Access-block Microsoft device-code grant for standard workforce accounts so kits cannot redeem a coached paste.
- MFA 2.0 cannot prevent device-code phishing, so block or limit device-code sign-in, for example with Conditional Access. Phishing-proof, device-bound authentication does not, by itself, stop tokens from issuing after a genuine-page MFA success.
- Treat Primary Refresh Token and refresh-token revoke as mandatory containment after any suspected device-code lure, not password reset alone.
- Hunt admin-panel style PRT fields and themed SharePoint or Zoom lures that open the real device-login URL, not only fake password domains.
Cloud identity keeps productizing convenience flows that another party can finish. The durable fix is to stop minting tokens for ceremonies a clipboard and a backend poll can steal.
FAQ
Would phishing-proof authentication alone have stopped BlackApex device-code token theft?
No. Phishing-proof device-bound authentication alone would not have stopped BlackApex-family initial access. According to LevelBlue SpiderLabs, victims finished MFA on Microsoft’s real device-login page while kit backends redeemed access and refresh tokens. Blocking or turning off device-code sign-in, for example with Conditional Access, is what closes that grant. MFA 2.0 cannot prevent device-code phishing. Primary Refresh Tokens already captured still need explicit revoke and session controls.
Why did MFA on Microsoft’s real page fail to protect mailboxes in the BlackApex flow?
MFA on Microsoft’s real page failed to protect mailboxes in the BlackApex flow because the victim completed MFA for a device code the kit already held. LevelBlue SpiderLabs reported that the victim completed MFA themselves, so the control was satisfied by the wrong person relative to who redeemed the tokens. The second factor never had to be typed into a spoofed form. Tokens still issued to the party polling the completed code.
Do passkeys alone stop Microsoft device-code phishing like BlackApex?
Passkeys alone are not shown in LevelBlue’s BlackApex reporting as blocking completion on the genuine Microsoft device-login page when device-code grant remains available. Passkeys are origin-bound at login through hardware-held keys and origin-bound signatures. Removing the device-code grant is what leaves no attacker-redeemable token path for a clipboard lure. Origin-bound passkeys do not change that, because the victim is already on the real Microsoft page.
What should defenders do about Primary Refresh Tokens after a BlackApex-style redemption?
Defenders should revoke access tokens, refresh tokens, and Primary Refresh Tokens immediately after a suspected BlackApex-style device-code redemption, and not rely on password reset alone. LevelBlue SpiderLabs reported PRT capture displayed in kit admin panels and used for mailbox access, with SSO-style persistence beyond a single session. Intake notes on this family also describe usability after password reset. Continuous access evaluation and shorter practical lifetimes support cleanup but do not replace prevention of the redemption path.