Safeguard
Vulnerability Analysis•

CVE-2026-75003: Remote Image Blocking Bypass in Roundcube

CVE-2026-75003 is a medium severity bypass of Roundcube remote image blocking using an unclosed url() in an SVG. Upgrade to 1.6.18 or 1.7.3.

Safeguard Security Team
3 min read

CVE-2026-75003 is a medium severity (CVSS 5.8) flaw in Roundcube Webmail. An unclosed url() in an SVG attribute can get past the remote image blocking, leading to information disclosure. Upgrade to 1.6.18 on the 1.6 line or 1.7.3 on the 1.7 line.

What's affected

Roundcube Webmail before 1.6.18, and 1.7.x before 1.7.3. The fixed versions are 1.6.18 and 1.7.3, and both release notes list the fix as a remote content blocking bypass via an unclosed url() in a FuncIRI attribute.

Both releases are described as security updates. The 1.6.18 notes list several other fixes, including SSRF filter bypasses, LDAP filter injection, Sieve script injection, an issue in the markasjunk plugin's cmd_learn driver, and IMAP command injection through mail search. So the reason to upgrade goes beyond this one entry.

How to check if you're exposed

Open Roundcube, then Settings, then About, to read the version. On the server, the version also appears in the release files you deployed.

If you run it from a distribution package, check the package version and the distribution's security notes, since fixes are often backported.

What to do now

  1. Upgrade to Roundcube 1.6.18 or 1.7.3, depending on your branch.
  2. Read the release notes for the other security fixes and for any configuration changes.
  3. Back up the database and configuration before upgrading, and run the update script as the upgrade notes direct.
  4. Check that remote content blocking is still on in your default settings.
  5. If you use the plugins named in the notes, such as markasjunk or managesieve, retest them after the upgrade.

Why it matters

The record maps this to CWE-669, incorrect resource transfer between spheres. Webmail clients block remote images by default, because loading an image from a server tells the sender that the message was opened, and from where. That blocking depends on the sanitizer finding every place an image address can hide, including inside SVG attributes.

An unclosed function style reference was not recognized by the filter, so a remote address could slip through. The impact is privacy leakage, which is why the score is medium, but the same release carries more serious fixes. The KEV flag is false.

Why remote content blocking matters

Mail clients that load remote images tell the sender that a message was read, when, and from what address. Spammers and targeted senders use that. Blocking remote content by default is a privacy control, and a bypass quietly turns it off for a certain class of message. The fix here is small, but the control it protects is one many users rely on without thinking about it.

Upgrade planning

Because the same release lists several other fixes, including server side request forgery filter bypasses and injection issues in the LDAP, Sieve and IMAP code, treat this as a normal security upgrade and plan it soon. Test with your plugins, since plugin code that touches the affected areas may behave differently. If you run Roundcube behind a corporate network, the SSRF fixes may matter to you more than this entry does.

The free in-browser dependency scan at gold.safeguard.sh/scan can show whether a project you maintain pulls in an affected package, and the Gold vulnerability directory keeps the record for this CVE.

Quick check

Is 1.6.18 enough on the 1.7 branch?

No. The 1.7 branch needs 1.7.3 or later. The 1.6 line is fixed in 1.6.18.

Sources

Last reviewed: 19 October 2026.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.