Roundcube 1.7.5 and 1.6.20 fix 20 security bugs, including two IMAP command injections via CRLF, a session unserialize object-injection primitive and a CSRF bypass; webmail admins should update
The Roundcube project shipped security updates 1.7.5 and 1.6.20 (1.6 LTS) on Oct 19. Each fixes 20 reported issues and the project recommends updating all production installs. Highlights from the release notes:
- IMAP command injection via CRLF in the
_uidparameter (mail/headers and mail/viewsource) and in the_flagparameter (mail/mark). - Object injection primitive in
rcube_session::unserialize()via "!"-prefixed session keys. - CSRF validation bypass in the mark-as-junk action, and deprecated action aliases that bypassed
disabled_actions. - Unsafe attachment content passed to ImageMagick on resize, plus an ImageMagick resource-limit bypass through the GD fallback.
- SSRF filter bypass via NAT64 (
64:ff9b::/96), remote-content-blocking bypasses (SVG SMIL, CSS escapes), CSS scoping bypass in HTML mail. - Several mail-header injections, XSS in password-driver error responses, the Modoboa password driver able to change another account's password, and an unsafe login-name path in SquirrelMail import.
The notes give no CVE IDs or severity scores, and no exploitation has been reported. Roundcube has a long record of being exploited by state actors (APT28 and Winter Vivern have both used Roundcube XSS bugs), and several of these flaws can be chained with message-borne input. Do now: update to 1.7.5 or 1.6.20. If you run 1.5.x, note this round shipped no 1.5 release, so plan the move to 1.6/1.7. If you can't patch today, disable the password plugin's Modoboa driver and the markasjunk plugin, and put the webmail behind SSO or a VPN.
Sources: Roundcube news: security updates 1.6.20 and 1.7.5 · GitHub release 1.7.5 · GitHub release 1.6.20