Skip to content

BlueKit (SpyCloud, Sep 15, 2026)

Tags

Summary

BlueKit is a full-featured subscription phishing-as-a-service kit documented by SpyCloud Labs on September 15, 2026 that SpyCloud tracks as the only PhaaS kit so far offering Browser-in-the-Middle (BitM), an evolution past adversary-in-the-middle (AiTM). Where AiTM proxies traffic between the victim's browser and the real login service, BitM loads the legitimate login page inside attacker-controlled infrastructure (a remote browser) and streams the rendered page to the victim, relaying live interactions back. The credential and MFA entry never touches a proxied HTTP session the way classic AiTM does; the victim is interacting with a real Microsoft/SSO page that just happens to be running somewhere else.

SpyCloud first recaptured BlueKit-stolen victim data on September 3, 2026, then collected ~1,000 attributable messages documenting session captures against 37 organizations, including U.S. and EU defense contractors, law firms, healthcare providers, multinational food distributors, and construction companies, arriving in two bursts (September 4 and September 10).

The marquee durable finding is the persistence step: in 38% of observed compromises BlueKit automatically enrolled its own TOTP authenticator on the victim's account after login. That requires driving a real browser inside the authenticated account (consistent with the BitM architecture, which has one anyway) and leaves the attacker with MFA-valid access that survives a password reset. SpyCloud's remediation guidance follows directly: incident response for session-theft phishing must audit registered MFA methods on the account, not just rotate the password and revoke active sessions, because a kicked session with an enrolled attacker authenticator simply comes back through the front door on the next login.

Kit capabilities (per SpyCloud)

  • Panel: subscriptions, phishing-site creation, automated domain purchasing and registration, campaign management with 80+ preconfigured templates (including device-code and BitM flows), delivery tooling (SMS sending + an AI assistant), and victim-log tracking in one interface.
  • Visitor qualification before anything malicious is served: custom CAPTCHAs, browser fingerprinting, WebRTC/STUN IP enumeration, device filtering, RAM/CPU checks, and IP-reputation filtering, all aimed at blocking security scanners and researchers (the same anti-scan posture this wiki notes on NovaCookies' Cloudflare gate and the RMM kit's disposable infra, now with hardware-floor checks).
  • Attribution-adjacent signals (suggestive, not proof): the kit's CIS-country blocking filter, advertising on Russian-language forums (XSS.is, Exploit.in, T1erOne, Lolzteam), and consistent Doraemon branding across panel/website/Telegram all suggest Russian-speaking developers.

Why BitM matters defensively

  • BitM is positioned as an evasion of the detectability of AiTM: defender tooling and user awareness have caught up to proxy-phishing (cookie-thepin, reverse-proxy tells, TLS anomalies on lookalike hosts). A remote-rendered genuine page defeats "the login page came from the real origin in your browser" checks, the origin is real, the browser isn't yours.
  • The attacker-enrolled TOTP step converts a one-time session theft into durable account ownership. Rotation workflows that don't enumerate authenticator registrations will "remediate" an account the attacker can still log into with a valid second factor.
  • Same intake window shape as other campaigns: two discrete data bursts a week apart = campaign scheduling, not continuous operation, useful for correlating victim exposure windows.
  • BlgCloud leak series (PwnForums): through August 2026, five apparently unaffiliated actor accounts coordinated to trickle out 15+ breaches stolen from customer instances of BlgCloud, a French business-management SaaS, every post titled "BLGCloud Leak #N," monetizing via unreleased-breach sales, selling "the access method," or pressuring BlgCloud itself. The pattern is the ShinyHunters extortion model (breach individual SaaS customer instances, pressure the provider) executed by a coordinating group that claims no shared brand, coordination without clout-seeking, which SpyCloud calls somewhat novel against the forum-billing backdrop.
  • LeakBase revival: after owner "Chucky" was arrested (March 2026) and the forum seized, LeakBase reappeared in late July 2026 under a new administrator holding a copy of the original backend database and administrative control of Chucky's old Telegram channels, mirroring the BreachForums→PwnForums copycat playbook (leaked backend + user base = instant forum). The new admin has released multiple unique datasets; SpyCloud assesses it is likely not a Russian MVD sting because a sting wouldn't keep publishing real data.

Defender actions

  1. Add MFA-method audit to every phishing-session IR runbook: on suspected session theft, list registered authenticator apps / TOTP seeds / FIDO2 keys on the account and remove unknown enrollments; password rotation + session revoke is incomplete remediation.
  2. Alert on MFA-enrollment events outside the user's normal change window (new TOTP enrollment is a rare, high-signal event in most tenants).
  3. Hunt the qualification-gate fingerprints: phishing landing pages that do WebRTC/STUN enumeration, RAM/CPU checks, or custom CAPTCHAs before serving content are PhaaS-qualifying visitors, treat passing such checks as an indicator the page is deliberate, not misconfig.
  4. Watch for BitM copycats: BlueKit is the first kit offering BitM in its template menu; the architecture is serviceable by other kit developers on the same timeline the AiTM proxies were commoditized.

Caveats

  • Victimology, counts (37 orgs, ~1,000 messages, 38% TOTP enrollment) are SpyCloud's own recaptured-data visibility, not a global census; burst timing reflects when stolen data was posted where SpyCloud watches.
  • "Only kit offering BitM" is SpyCloud's tracking statement as of September 2026.
  • Russian-speaking origin is inferred from forum presence, branding, and CIS filters, suggestively-patterned, not attributed to any named group.

Sources