1-minute read of a 5 min report

Zimbra Mail Clusters Face Secret Theft After Server Intrusions

By

1-minute read

What a defender needs to know

Microsoft’s September 30 investigation follows exploitation of CVE-2026-73570 beyond the first Zimbra mail server. With optional zimbra-snmp installed and SNMP notifications enabled, crafted mail reaches a swatchdog-to-snmptrap shell path and runs commands as zimbra. Intruders then planted JSP shells, gained root access in one documented mailbox-manager/PAM sequence, collected zimbraPreAuthKey and other authentication material, and used /opt/zimbra/.ssh/zimbra_identity to reach mailbox peers. Microsoft’s account combines separate confirmed compromises; individual hosts showed different stages.

Zimbra fixed the flaw in 10.1.20. Check that release and the optional package/notification state on each internet-facing node. If patching must wait, remove zimbra-snmp, turn off SNMP notifications and restrict SNMP and SMTP to trusted hosts as Microsoft advises. For possible compromise, correlate unexpected shell children of swatchdog or snmptrap with JSP files in Jetty or mailboxd paths, /etc/systemd/system/zimlog.service, changes to /etc/pam.d/sudo or sudoers, and peer SSH activity. Contain affected nodes, examine their peers and rotate every domain zimbraPreAuthKey; review other exposed credentials and token material. Confirm the installed release and compare web files, units, sudoers and SSH events with known-good state. Use archived telemetry for July and August because the source’s 30-day Defender XDR raw-event window no longer reaches the early probing.

Primary sources