Device code phishing jumps 37x in 2026 as EvilTokens and copycat kits spread
Push Security saw device code phishing pages rise 37.5x in 2026 as kits like EvilTokens hijack OAuth tokens and sidestep MFA. How to block and detect the flow.

Device code phishing tricks users into approving an attacker's sign-in on a real login page, and in 2026 it went mainstream. Push Security reported in April that device code phishing pages it detected had risen 37.5 times so far this year, driven by ready-made phishing-as-a-service kits. Because the victim authenticates on the genuine identity provider, passwords, MFA and even passkeys do not stop it. Developer-heavy organisations are in an awkward spot, since tools like the GitHub CLI rely on the same flow.
What happened
On 4 April 2026, Push Security researcher Luke Jennings wrote that the number of device code phishing pages Push detected had reached 37.5x, up from a 15x increase it had observed at the start of March. Push links the surge to EvilTokens, a phishing-as-a-service platform documented by Sekoia that packages the technique for low-skilled criminals, and identifies more than a dozen competing kits. Names include SHAREFILE, Kali365 (CLURE), LINKID, Device Code Lab (formerly AUTHOV), DOCUPOLL, FLOW_TOKEN, PAPRIKA and DCSTATUS; a May update added Venom, Tycoon2FA and CYB3R.
Lures imitate SharePoint, Teams, DocuSign, Adobe, Citrix ShareFile and HR notices, use anti-bot checks and are hosted on legitimate platforms such as Cloudflare Workers, GitHub Pages, AWS S3 and Microsoft PowerApps. The campaigns Push and BleepingComputer describe focus on Microsoft 365 accounts.
Who is affected
- Microsoft Entra ID / Microsoft 365 tenants where the device code flow is allowed, which is the default unless blocked by Conditional Access.
- Any OAuth provider that supports the Device Authorization Grant (RFC 8628). The technique is not Microsoft-specific, though current kits target Microsoft.
- Users with an active session, who can approve the request with a click and no fresh MFA prompt.
What it means for developers and engineering teams
The OAuth 2.0 Device Authorization Grant exists for devices that cannot easily show a browser, like smart TVs and CLIs. The client calls the provider's device authorization endpoint and receives a device_code, a short user_code and a verification URL. The user enters the code on another device, signs in and approves. Meanwhile the client polls the token endpoint and, once approved, receives access and often refresh tokens.
The flaw attackers exploit is that the first request is unauthenticated and nothing ties the code to the person who started it. The attacker requests a code, sends it to the victim with a plausible pretext ("enter this code to view the document"), and the victim approves it on the real Microsoft page. The attacker's poller then gets the tokens. Modern kits request fresh codes on demand through an API, sidestepping the roughly 15-minute code expiry. In Microsoft environments, Push notes the stolen tokens can be extended into a Primary Refresh Token or used across Microsoft apps via Family of Client IDs.
If you build a CLI or device app on this flow, design the approval screen so users can see what they are approving: show the application name, requested scopes and, where possible, the location of the requesting device. Keep scopes minimal and code lifetimes short.
What to do now
- Block device code flow by default in Microsoft Entra Conditional Access: target all users, set the Authentication flows condition to Device code flow, and grant Block access. Run it in report-only mode first to find legitimate use.
- Add narrow exceptions for the users, devices or apps that genuinely need it, such as meeting-room devices or specific developer tooling.
- Hunt in sign-in logs for device code authentications. In Microsoft Sentinel:
Look for users who never use the flow, and for mismatches between the approval IP and later activity.SigninLogs | where AuthenticationProtocol == "deviceCode" | project TimeGenerated, UserPrincipalName, AppDisplayName, IPAddress, Location - Revoke sessions and refresh tokens for any account that approved an unexpected device code.
- Train users that no legitimate document share asks them to type a code into a Microsoft login page.
- Do not assume blocking ends the risk. Push notes related techniques such as ConsentFix abuse the authorization code flow to similar effect.