Russian state actor LAUNDRY BEAR weaponised a Zimbra webmail zero-click (CVE-2025-66376) for mailbox exfiltration, now exposed in a 16-nation joint advisory
16-nation advisory: Russia's LAUNDRY BEAR exfiltrates government mail through a view-based Zimbra exploit, and patching alone does not evict it
Defender actions
- Confirm every on-prem Zimbra Collaboration Suite instance is on ZCS ≥ 10.0.18 or ≥ 10.1.13 specifically for CVE-2025-66376; this is a distinct patch from the ZCS 10.1.19 and 10.1.20 fixes tracked earlier this month.
- On any ZCS account suspected of exposure, revoke all application-specific passwords and regenerate 2FA scratch codes, then disable IMAP where it is not required, patching stops the next crafted email but does not revoke access an earlier one already established.
- On any Zimbra account touched by the CVE-2025-66376 campaign, revoke application-specific passwords via the admin console, not just reset the user password: the attacker-created 'ZimbraWeb' app-password survives a password reset and the patch and grants standalone IMAP/POP3/SMTP mailbox access.
Analysis
Background. LAUNDRY BEAR; named by the Netherlands' AIVD/MIVD in May 2025, tracked as Void Blizzard by Microsoft, CL-STA-1114 by Unit 42 and TA488 (formerly UNK_PitStop) by Proofpoint, is a Russian state-supported espionage actor that, until this campaign, relied on password spraying, adversary-in-the-middle credential phishing (a modified Evilginx) and pass-the-cookie against Microsoft Exchange and cloud mail (CISA, 2026-07-23). The 2026-07-23 joint advisory AA26-204A, co-sealed by security and intelligence agencies from sixteen US, Five Eyes and EU-member nations including ANSSI/DGSI (France), AISE/AISI (Italy), AIVD/MIVD (Netherlands) and NCSC-UK, documents its shift to a genuinely technical exploit and assesses the covert, non-extortion nature of the activity as "almost certainly" Russian-government-backed espionage (CISA, 2026-07-23).
Since at least July 2025 the actor has exploited CVE-2025-66376, a stored cross-site-scripting flaw in the Zimbra Collaboration Suite (ZCS) Classic Web Client caused by improper sanitisation of CSS @import directives in HTML email (Unit 42, 2026-07-23). A crafted message smuggles a Base64-encoded, XOR-obfuscated JavaScript payload inside an SVG element's onload attribute; the script executes inside the victim's authenticated webmail session the moment the message is rendered. CISA describes this precisely as view-based; it "only requires a user to view a malicious email within a vulnerable version of the webmail service" (CISA, 2026-07-23), while NCSC-UK and Unit 42 call it zero-click; with preview-pane rendering the practical effect is the same, and there is no link to click or attachment to open. It was a true zero-day when first used, and Synacor did not patch it until ZCS 10.0.18 / 10.1.13 in November 2025, yet Unit 42 confirms attackers "continue to actively target unpatched ZCS instances" today (Unit 42, 2026-07-23). This is a third, distinct ZCS Classic Web Client flaw from the two unrelated July 2026 issues this store already tracks (the no-CVE code-exec fixed in 10.1.19 and the SNMP command-injection RCE fixed in 10.1.20).
Kill chain. On render, the payload ("Ulej", Russian for beehive) pulls the session CSRF token from browser localStorage and drives a 12-stage asynchronous collection routine over the Zimbra SOAP API: it retrieves the victim's last ~90 days of mail, brute-forces the organisation's Global Address List through repeated two-character SearchGalRequest batches (roughly twenty batches of ~77 queries), harvests GetScratchCodesRequest 2FA backup codes, and (the persistence pivot) issues CreateAppSpecificPasswordRequest to mint an IMAP application passcode (observed named ZimbraWeb) that bypasses ZCS's lack of native 2FA on IMAP, then flips zimbraPrefImapEnabled to true (CISA, 2026-07-23). Because that application passcode is a legitimate credential, it survives both the November patch and any subsequent password reset, as The Hacker News put it in its analysis, patching "stops the next crafted email from running. It does not revoke what the last one left behind" (The Hacker News, 2026-07-23). Exfiltration runs over both Base32-encoded DNS A-record queries and HTTPS to a Dockerised "Flowerbed" collection stack (Catcher/Certbot/Nginx/Gardener containers, Let's Encrypt via Cloudflare DNS challenge) on short-lived VPS infrastructure reached over Mullvad VPN; CISA notes the simplistic Flowerbed codebase shows indications of AI-assisted development. Since ~November 2025 the actor has also sent lures from already-compromised victim mailboxes, defeating sender-reputation filtering (CISA, 2026-07-23). Targeting spans the Defence Industrial Base, federal and local government, education, energy, law enforcement, media and NGOs across NATO states, with Ukraine used as an earlier testbed (NCSC-UK, 2026-07-23).
Cited evidence
Unlike traditional phishing campaigns that persuade a user into taking an action, such as clicking a link or opening a file, LAUNDRY BEAR's latest campaign leverages a view-based exploit that only requires a user to view a malicious email within a vulnerable version of the webmail service.
A patch for CVE-2025-66376 was released for both 10.1.13 and 10.0.18 versions of ZCS.
Threat actors continue to actively target unpatched ZCS instances using CVE-2025-66376.
the sanitizer fails to recognize it as executable markup, while the browser successfully reconstructs <svg onload="eval(atob('…'))"> and executes it.
Proofpoint has not observed TA458 using CVE-2025-66376, despite the group's regular access to webmail XSS zero-days.
Updates1
The prior entry covered the 16-nation joint advisory (AA26-204A) on LAUNDRY BEAR's (TA488 / Void Blizzard) zero-click Zimbra campaign exploiting CVE-2025-66376. Proofpoint's own two-part writeup adds mechanics the advisory did not detail (Proofpoint, 2026-07-23).
The exploit defeats Zimbra's client-side HTML sanitizer by tag-splitting: a fake CSS @import directive is fragmented across HTML tags so that the sanitizer strips each @import sequence individually, but "the browser successfully reconstructs <svg onload="eval(atob('…'))"> and executes it", a sanitizer-bypass-by-reassembly, where the surviving characters rejoin into valid executing script (Proofpoint, 2026-07-23). The resulting payload (which Proofpoint attributes to LAUNDRY BEAR's post-exploitation tooling and names ZimReaper) steals CSRF tokens, browser-autofill credentials, 2FA scratch codes and the Zimbra version/URL, exfiltrating tokens via DNS tunneling (Base32-encoded subdomains carrying a session id and a plaintext token-type label) and mail archives via HTTP POST.
The operationally important delta is persistence: after exploitation ZimReaper issues a CreateAppSpecificPasswordRequest to mint a Zimbra application-specific password labelled "ZimbraWeb" that bypasses 2FA and grants IMAP/POP3/SMTP access, then exfiltrates it via DNS for direct mailbox access (Proofpoint, 2026-07-23). Because that credential is independent of the primary account password, it survives a password reset and the CVE-2025-66376 patch alike.
Sources5
Revision history
- Published 2026-07-24T0409Z-intel
- Update 2026-07-25T0409Z-intel
Proofpoint's writeup of the LAUNDRY BEAR (TA488/Void Blizzard) Zimbra CVE-2025-66376 campaign adds the technical mechanics the 16-nation joint advisory did not spell out: the CSS-@import sanitizer-bypass-by- reassembly that reconstructs an executing <svg onload=eval(atob(...))>, DNS-tunnelled exfiltration, and a persistent "ZimbraWeb" application-specific password created via the SOAP API that survives both a user password reset and the CVE-2025-66376 patch.
Changed: actions affected_products evidence references tags techniques body
AI-generated · no human review · this permalink is the shareable record for the finding · verify operationally critical claims against the linked primary source.