Russian State-Sponsored Espionage Campaign Exploits Critical Zimbra Zero-Click Vulnerability to Target Western Government and Defense Organizations

The National Security Agency (NSA), in coordination with the Cybersecurity and Infrastructure Security Agency (CISA), the Federal Bureau of Investigation (FBI), and international partners from the United Kingdom, Canada, Australia, and New Zealand, has issued a comprehensive joint cybersecurity advisory regarding a prolonged espionage campaign conducted by Russian state-supported actors. This campaign, which has been active since at least July 2025, leveraged a previously unknown vulnerability in the Zimbra Collaboration Suite (ZCS) to infiltrate and monitor the private communications of government agencies, military contractors, and critical infrastructure entities across the West.
The exploitation centered on a stored cross-site scripting (XSS) vulnerability, now tracked as CVE-2025-66376, which resides within Zimbra’s Classic Web Client interface. According to forensic evidence gathered by cybersecurity firms Palo Alto Networks’ Unit 42 and Proofpoint, the threat actors—identified by various names including TA488, LAUNDRY BEAR, and Void Blizzard—successfully operated for at least five months using this flaw as a zero-day exploit before a patch was made available by the vendor.
Technical Analysis of CVE-2025-66376 and the "Zero-Click" Vector
The vulnerability is particularly dangerous because it facilitates what security researchers call a "view-based" or "zero-click" exploit. Unlike traditional phishing attacks that require a user to click a suspicious link or download a malicious attachment, this exploit is triggered the moment a victim simply opens or previews a specially crafted HTML email within a vulnerable Zimbra client.
The technical mechanism involves a sophisticated bypass of Zimbra’s built-in HTML sanitization engine. The attackers utilized a technique described as "tag-splitting," where executable JavaScript code is broken into fragments and hidden within CSS @import directives and HTML comments. To a standard security filter, these fragments appear as harmless, non-executable styling instructions. However, once the email is rendered by the victim’s browser, the Zimbra sanitizer strips away the deceptive @import sequences, causing the remaining characters to coalesce into a functional <svg onload> tag.

This reconstructed tag executes a JavaScript payload within the context of the user’s authenticated webmail session. Because the script runs within the active session, it inherits all the permissions and access rights of the logged-in user, allowing the attackers to interact with the Zimbra API as if they were the legitimate account holder.
There has been a notable discrepancy in the severity scoring of CVE-2025-66376. The National Vulnerability Database (NVD) assigned the flaw a CVSS score of 6.1, categorizing it as requiring user interaction. Conversely, MITRE and Unit 42 have argued for a higher score of 7.2, classifying it as a zero-click vulnerability because the "interaction" of simply viewing an email is considered passive and unavoidable in a standard workflow.
The ZimReaper Payload and Data Exfiltration
Once the exploit successfully triggers, it deploys a sophisticated JavaScript-based payload, which Proofpoint has dubbed "ZimReaper." This malware is designed for rapid and comprehensive data harvesting. The primary objectives of the payload include:
- Email Theft: The script automatically archives and exfiltrates the last 90 days of the victim’s email correspondence.
- Directory Harvesting: It performs a brute-force query of the Global Address List (GAL). By iterating through two-character combinations, the payload can reconstruct the organization’s entire internal directory, providing the attackers with a roadmap for future targeting.
- Credential Harvesting: ZimReaper targets the browser’s autofill passwords and extracts the user’s Cross-Site Request Forgery (CSRF) tokens.
- MFA Circumvention: The payload attempts to pull two-factor authentication (2FA) recovery or "scratch" codes kept within the Zimbra platform.
- Persistence Creation: One of the most critical functions of the payload is the unauthorized generation of an app-specific password, often named "ZimbraWeb." This allows the attackers to maintain persistent access to the mailbox via IMAP, POP3, or SMTP protocols, effectively bypassing multi-factor authentication requirements even if the user later changes their primary account password.
The exfiltration process is equally stealthy. Rather than using standard HTTP POST requests, which are more easily monitored by firewalls, ZimReaper often encodes the stolen data into DNS queries directed at actor-controlled infrastructure. This technique, known as DNS tunneling, allows the data to blend in with legitimate network traffic, making detection significantly more difficult for traditional security operations centers (SOCs).
Chronology of the Espionage Campaign
The timeline of this campaign reveals a calculated and patient approach by the Russian intelligence services.

- July 2025: Initial exploitation begins. TA488 starts targeting Western government entities using CVE-2025-66376 as a zero-day vulnerability.
- July – October 2025: The campaign expands to include NATO member states, defense industrial base (DIB) entities, and scientific research organizations.
- November 6, 2025: Zimbra releases a patch (version 10.0.18 and 10.1.13) to address the XSS flaw. However, many organizations remain unpatched or fail to conduct necessary post-patch account audits.
- December 31, 2025: Zimbra Collaboration 10.0 reaches its End of Life (EOL), leaving organizations on this version without further security updates.
- January 2026: Security firm Seqrite analyzes a compromise at a Ukrainian state hydrology agency, identifying similar tactics and the modification of IMAP settings to ensure long-term access.
- February 2026: Proofpoint notes a decline in TA488 activity as the actor begins tearing down its command-and-control (C2) infrastructure following public disclosures.
- March 18, 2026: CISA officially adds CVE-2025-66376 to its Known Exploited Vulnerabilities (KEV) catalog, mandating that federal agencies apply patches immediately.
- July 2026: Continued activity is observed. Unit 42 reports that various threat actors continue to scan for and exploit unpatched Zimbra instances, despite the availability of fixes for nearly eight months.
Targeted Sectors and Geopolitical Context
The scope of the campaign suggests a high-level strategic intelligence-gathering mission. Unit 42’s analysis identified at least nine distinct command-and-control IP addresses and domains, with each server remaining active for approximately 35 days.
The targets were not chosen at random. The campaign focused heavily on:
- NATO Member States: Particularly those involved in providing strategic or military support to Ukraine.
- Ukraine: Targeting government agencies, including those involved in hydrology and logistics.
- United States Infrastructure: Specifically defense industrial base entities and nuclear installations.
- The Commonwealth of Independent States (CIS) and Africa: Targeting diplomatic and financial organizations to monitor regional geopolitical shifts.
The involvement of LAUNDRY BEAR (also known as APT28 or Fancy Bear) underscores the state-sponsored nature of the threat. While some intelligence agencies, such as those in the Netherlands, treat LAUNDRY BEAR and APT28 as distinct entities, the consensus among Western partners is that both operate under the umbrella of Russian military intelligence (the GRU).
Broader Impact and Industry Reactions
The exploitation of Zimbra highlights a recurring theme in modern cyber-espionage: the targeting of secondary or open-source enterprise software that may not receive the same level of security scrutiny as Microsoft Exchange or Google Workspace.
"State-sponsored actors are increasingly moving toward zero-click exploits because they bypass the most common point of failure: the user," said one senior cybersecurity analyst. "When the act of simply checking your inbox becomes a security risk, the entire trust model of digital communication is challenged."

Zimbra has urged its customers to move to the latest supported versions (currently 10.1.20) to defend against not only CVE-2025-66376 but also four subsequent XSS flaws discovered in the Classic Web Client. The company emphasized that while patching is essential, it is only the first step in remediation.
Remediation and Defensive Recommendations
For organizations utilizing Zimbra Collaboration Suite, the joint advisory stresses that applying the software update does not automatically remediate a compromise. If an account was accessed by the ZimReaper payload, the attackers may have already established persistent backdoors.
Defenders are advised to implement the following protocol:
- Immediate Patching: Update all Zimbra deployments to at least version 10.1.13. Organizations still on the EOL version 10.0 must migrate to a supported 10.1 build.
- Account Auditing: For any mailbox suspected of viewing a malicious email, administrators must:
- Reset the user’s primary password.
- Invalidate all active web sessions.
- Review and revoke any "App-Specific Passwords" created during the window of vulnerability.
- Regenerate all 2FA recovery and scratch codes.
- Configuration Check: Verify that
zimbraPrefImapEnabledand other protocol settings have not been unauthorizedly modified to allow external access. - Log Analysis: Search for the fragmented
@importpattern within email logs using YARA rules provided by Proofpoint and CISA. - Network Monitoring: Monitor DNS traffic for unusual patterns or queries to known C2 domains associated with TA488.
The persistence of these Russian-linked groups suggests that even as this specific campaign winds down, the tactics used will likely be adapted for other webmail platforms. The NSA and its partners continue to monitor the situation, warning that the "ZimbraWeb" persistence mechanism is a particularly potent tool that could allow attackers to remain inside Western networks for years if left undetected.







