CVE-2026-73570, Zimbra Collaboration: a pre-auth command injection patched without a CVE in July is exploited, with probing before disclosure and root escalation, secret theft and cluster-wide movement observed
Fixed on 20 July, identified on 13 August, probed from 28 July: a CVE-keyed patch process never saw this one coming
Defender actions
- Upgrade every Zimbra Collaboration host to 10.1.21 (10.1.20 is the minimum for CVE-2026-73570); where the upgrade cannot happen at once, remove the zimbra-snmp package or disable SNMP notifications and restrict SNMP and SMTP to trusted hosts.
- On every host that ran a build before 10.1.20 with SNMP notifications enabled at any time since 2026-07-28, upgrading is not enough: rotate all domain zimbraPreAuthKey values and run the compromise check on every mailbox node, covering the zimbra user's crontab and /opt/zimbra/.ssh/authorized_keys, JSP or stray text files under /opt/zimbra/jetty_base/webapps/zimbra/public, droppers in /tmp and /var/tmp, and outbound connections from the mail server; where web shells or an LDAP dump turn up, rotate every mailbox and administrative credential.
Analysis
Zimbra's own security-advisory table records the fix for CVE-2026-73570 as "Fixed a command injection vulnerability in the SNMP monitoring component when SNMP notifications are enabled", shipped in release 10.1.20 (Zimbra, 2026-08-13). That release went out on 20 July 2026 (Zimbra, 2026-07-20) carrying nine fixes, and at the time none of them had been flagged as actively exploited; the vendor's stated position was that "in line with industry best practices, information disclosure is limited for security vulnerability fixes" (The Hacker News, 2026-07-21). The identifier arrived 24 days later, on 13 August, and the ENISA record describes the mechanism in full: because untrusted input is not properly sanitised during SNMP notification processing, "an unauthenticated attacker can send specially crafted SMTP requests that may result in execution of arbitrary operating system commands as the Zimbra user" (ENISA EU Vulnerability Database, 2026-08-13). ENISA scores it 8.9 with high attack complexity (ENISA EU Vulnerability Database, 2026-08-13), and the flaw applies only to deployments where the optional zimbra-snmp package is installed and SNMP notifications are enabled.
On 19 August CERT-FR issued its own advisory for the Zimbra bulletin and stated plainly that ENISA records CVE-2026-73570 as actively exploited (CERT-FR, 2026-08-19). ENISA's record lists it in the EU KEV catalog from 18 August (ENISA EU Vulnerability Database, 2026-08-13). What makes this worth an out-of-band look rather than a place in the next patch window is the sequence rather than the score: the code was fixed in July with no identifier attached, so an estate that drives its patching from CVE feeds, scanner signatures or an SBOM pipeline had nothing to match against for 24 days, and the flaw only became visible to those processes five days before it was recorded as exploited. Anyone who upgraded to 10.1.20 in July for unrelated reasons is already covered and does not know it; anyone who deferred is now unpatched against a flaw with a published exploitation status.
Cited evidence
ENISA indicates that vulnerability CVE-2026-73570 is being actively exploited. (translated from French)
Fixed a command injection vulnerability in the SNMP monitoring component when SNMP notifications are enabled.
none of the identified vulnerabilities have been flagged as actively exploited
Between July 28 and August 7, after a fix became available on July 20 but before public disclosure on August 13, Microsoft observed two distinct out-of-band scanning tools probing the vulnerable injection point.
Rotate all domain zimbraPreAuthKey values and review systemd units for unexpected ownership, enablement, or timestamp changes
In two separate events, exploitation of CVE-2026-73570 culminated in the release of ransomware against the target servers. (translated from Italian)
No evidence emerged of successful intrusions against regularly updated instances. (translated from Italian)
several of the public proofs of concept (PoCs) we tested write a payload that Swatchdog never matches, so a failed test looks like a patched server.
Updates3
Microsoft Threat Intelligence published first-hand analysis on 2026-09-30 of CVE-2026-73570 exploitation in more than one region and industry, naming no actor (Microsoft Threat Intelligence, 2026-09-30). It saw two out-of-band scanning tools probing the injection point between 2026-07-28 and 2026-08-07, after the fix and before the 2026-08-13 disclosure (Microsoft Threat Intelligence, 2026-09-30). CISA's KEV catalog lists the CVE, added 2026-08-21 (CISA KEV, 2026-08-21).
After command execution as the zimbra account, Microsoft observed, across its cases and not necessarily on every host, JSP web shells written to publicly served directories after temporarily opening write permission, with copies on peer mailbox nodes, and reverse shells built from a named pipe and openssl s_client; a privilege escalation that used a symlinked log file to take ownership of the sudo PAM configuration, added a pam_exec hook and created a NOPASSWD sudoers entry for the zimbra account; a systemd unit named like a Zimbra logging component with timestamps matched to existing services; theft of Zimbra's service credentials with zmlocalconfig -s, followed by authenticated LDAP queries for the pre-authentication key, the auth-token key and the two-factor secret attribute; and SSH and rsync movement across the cluster with Zimbra's own SSH identity (Microsoft Threat Intelligence, 2026-09-30). Microsoft's remediation is to upgrade to 10.1.20 or later, remove the zimbra-snmp package or disable SNMP notifications and restrict SNMP and SMTP to trusted hosts, rotate all domain zimbraPreAuthKey values, review systemd units, and hunt for JSP files on every mailbox node (Microsoft Threat Intelligence, 2026-09-30).
Zimbra's 10.1.21 release of 2026-09-24 also fixes unauthenticated prediction of password-recovery codes that could reset a user's password, WebDAV acceptance of pre-MFA tokens, and OnlyOffice-integration flaws including file write to remote code execution (Zimbra, 2026-09-24); the table lists no CVE or score for the recovery-code and WebDAV fixes (Zimbra, 2026-08-13), and NCSC Switzerland published an advisory for the release on 2026-10-01 with exploitation status unknown (NCSC Switzerland, 2026-10-01).
CSIRT Italia reports, in bulletin BL01/261005/CSIRT-ITA, multiple compromises of internet-exposed Zimbra servers handled in the Italian national context, mostly through CVE-2026-73570 and to a lesser extent through older flaws on unpatched or end-of-life instances (ACN / CSIRT Italia, 2026-10-05). In two CVE-2026-73570 cases exploitation ended in ransomware: on an 8.7.11 instance the attacker opened an interactive reverse shell and irreversibly encrypted many files on one node, and in a second case of unspecified version an ELock-family ransomware executable was dropped; both servers had the optional zimbra-snmp package, and the evidence does not tie the two cases to one actor or payload (ACN / CSIRT Italia, 2026-10-05). Three further cases installed cryptominers, one on 10.1.19 with a crontab entry for the zimbra user that ran a payload from a temporary directory (ACN / CSIRT Italia, 2026-10-05). In one case the exploit fetched a payload from external staging infrastructure and ran a Perl script directly in memory without writing files, with an attempted connection to a presumed IRC command-and-control server and no post-compromise activity found (ACN / CSIRT Italia, 2026-10-05). In two cases web shells were planted in the directories the web service exposes: on a 10.1.4 instance the attackers left JSP web shells, including ones managed with Behinder, and exfiltrated the LDAP database, the mailbox list and the credential hashes, and on another instance several JSP web shells took mail down for 48 hours for more than 6,000 mailboxes with no data theft documented (ACN / CSIRT Italia, 2026-10-05). The two cases share five command-and-control addresses and similar JSP constructs, but ACN says the evidence does not let it attribute both to the same actor (ACN / CSIRT Italia, 2026-10-05). ACN also describes a CentOS server turned into an SSH brute-force source after a write test in the Jetty public directory, with CVE-2026-73570 as the presumed but unconfirmed entry, and an end-of-life server whose mailbox passwords were all reset by the attacker through an undetermined vector (ACN / CSIRT Italia, 2026-10-05). Older flaws (CVE-2024-45519, CVE-2022-41352, CVE-2022-27925, CVE-2022-37042 and CVE-2023-38750) account for the remaining cases on unpatched or unsupported lines, including a further ransomware case on an 8.8.15 instance (ACN / CSIRT Italia, 2026-10-05).
ACN says nearly all affected systems lacked applicable patches or ran unsupported branches, that in several cases the attackers persisted without root by staying in the zimbra application user's context, and that it found no evidence of successful intrusions on regularly updated instances (ACN / CSIRT Italia, 2026-10-05). Its mitigations are to update, to disable or remove the zimbra-snmp package where patching has to wait, to inspect the zimbra user's crontab and /opt/zimbra/.ssh/authorized_keys, to look for SNMP exploit scripts under /opt/zimbra/data/tmp and for JSP files or stray text files in the Jetty public directory, to check /tmp and /var/tmp for droppers, miners and ransomware payloads, to rotate all mailbox and administrative credentials where web shells or anomalies are found, to keep the administrative port TCP 7071 and unneeded SOAP interfaces off the public internet, and to filter egress from the Zimbra servers, including IRC and TCP 8801 (ACN / CSIRT Italia, 2026-10-05).
Elastic Security Labs reproduced the exploit chain in its lab and describes it stage by stage: an SMTP client sends a recipient address with an unclosed angle bracket, Postfix logs it to /var/log/zimbra.log as an illegal-address warning, Zimbra's bundled Swatchdog daemon matches a log line of the form "Service status change: <host> <service> changed from stopped to running", and the notification routine places the service field into a Perl backtick string that /bin/sh runs, giving code execution as the zimbra user, who can read every mailbox and send as any local identity (Elastic Security Labs, 2026-10-23). Its practical finding is that several public proofs of concept write a payload that Swatchdog never matches, so a failed test looks like a patched server; Elastic's own proof of concept proves execution with an out-of-band callback that changes nothing on disk, and it says upgrading to 10.1.20 or later is the only complete fix (Elastic Security Labs, 2026-10-23).
Its detection is a process sequence on one host within 30 seconds, a shell started by the Perl Swatchdog script with snmptrap in its command line followed by any child of that shell other than snmptrap; on a healthy host the notification shell starts only snmptrap, so any other child is the discriminator (Elastic Security Labs, 2026-10-23). Where exploitation is found, Elastic lists standard indicators of remote code execution, phishing mail originating from the server, compromise of local databases, signing keys and LDAP exports, and less direct surfaces beyond the mail store (Elastic Security Labs, 2026-10-23).
Sources11
Revision history
- Published 2026-08-20T0409Z-intel
- Update 2026-10-02T0404Z-intel
Microsoft Threat Intelligence published first-hand analysis of confirmed intrusions: probing of the injection point from 2026-07-28, before the 2026-08-13 disclosure, then web shells, root escalation, theft of Zimbra's authentication keys and cluster-wide movement, with the injection signature to hunt for. The 10.1.21 release, which also fixes unauthenticated password-recovery code prediction, and NCSC-CH's advisory of 2026-10-01 are added. CISA's KEV listing of 2026-08-21, not recorded before, is added; the 10.1.20 release date is corrected to 2026-07-20, the EPSS score is refreshed to 0.117 (FIRST, 2026-10-01), ENISA's 18 August date is its EU KEV listing, and the actions now ask for a compromise check as well as the upgrade.
Changed: title headline summary tags techniques cves sources evidence sourcing_note classification actions update_of body
- Update 2026-10-06T0405Z-intel
CSIRT Italia's bulletin of 2026-10-05 reports compromises of exposed Zimbra servers through CVE-2026-73570 that ended in ELock-family ransomware, cryptomining, web shells with theft of the LDAP database and credential hashes, and a 48-hour mail outage for more than 6,000 mailboxes, and found no intrusion on regularly updated instances. Its host checks (the zimbra user's crontab and authorized_keys, JSP files in the Jetty public directory, droppers in temporary directories, egress) are added to the compromise check.
Changed: summary tags entities techniques sources evidence sourcing_note actions body
- Improvement 2026-10-24T0255Z-intel
Elastic Security Labs reproduced the chain and found that several public proofs of concept never trigger it, so a failed test does not show a host is patched; its process-lineage detection for the Swatchdog notification path is added to the detection guidance.
Changed: body evidence sources sourcing_note
AI-generated · no human review · this permalink is the shareable record for the finding · verify operationally critical claims against the linked primary source.