Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-114716 Overview
CVE-2026-114716 is a mail header injection vulnerability in Roundcube Webmail affecting versions before 1.6.20 and 1.7.x before 1.7.5. The flaw resides in rcmail_sendmail::extract_inline_images(), which uses the media type extracted from an inline image data URI without validation. A remote authenticated attacker can inject arbitrary MIME part headers into outgoing messages by embedding CRLF sequences in the media type field of a data URI. The issue is classified as CWE-99: Improper Control of Resource Identifiers.
Critical Impact
Authenticated users can forge MIME headers in outbound mail, enabling message tampering, spoofed content parts, and potential abuse of downstream mail processing systems.
Affected Products
- Roundcube Webmail versions prior to 1.6.20
- Roundcube Webmail 1.7.x versions prior to 1.7.5
- Downstream distributions packaging vulnerable Roundcube releases (see Debian Bug Report #1151279)
Discovery Timeline
- 2026-10-19 - Roundcube publishes Security Updates Announcement 1.6.20 and 1.7.5
- 2026-10-22 - CVE-2026-114716 published to NVD
- 2026-10-22 - Last updated in NVD database
Technical Details for CVE-2026-114716
Vulnerability Analysis
Roundcube's compose pipeline extracts inline images referenced through data: URIs when building outgoing messages. The rcmail_sendmail::extract_inline_images() function in program/include/rcmail_sendmail.php parses the data URI with a regular expression and reuses the captured media type when constructing the attached MIME part. Because the captured value is placed directly into header fields without syntactic validation, carriage return and line feed characters within the media type break out of the header context. An authenticated attacker can leverage this to insert additional MIME headers, add rogue message parts, or manipulate Content-Type boundaries in the final SMTP payload.
Root Cause
The root cause is missing input validation on an untrusted substring that is later interpolated into mail headers. The pre-patch code only tested whether $mime_type was empty before falling back to automatic detection. It never checked that the string matched the grammar for a valid MIME type, allowing CRLF and other illegal characters through to the message builder.
Attack Vector
Exploitation requires an authenticated Roundcube session. The attacker composes a message containing HTML with an inline image reference such as <img src="data:image/png;...CRLF-INJECTED-HEADERS...;base64,...">. When Roundcube processes the draft for sending, the malicious media type is copied into MIME headers, injecting attacker-controlled header fields into the outgoing message.
// Patched code in program/include/rcmail_sendmail.php
$hash = md5($data) . '@' . $domain;
$mime_type = $matches[2][$idx][0];
// Before: only an emptiness check
// if (empty($mime_type)) {
// After: additionally validates MIME type syntax
if (empty($mime_type) || !rcube_mime::is_mimetype_valid($mime_type)) {
$mime_type = rcube_mime::image_content_type($data);
}
Source: Roundcube commit 7f3cc9c and commit adb9b9b. The patch introduces a call to rcube_mime::is_mimetype_valid() so any media type containing CRLF or other invalid characters is discarded in favor of auto-detection.
Detection Methods for CVE-2026-114716
Indicators of Compromise
- Outbound messages originating from Roundcube with duplicate or unexpected MIME headers in multipart boundaries.
- Draft messages containing HTML <img> tags with data: URIs where the media type field contains %0d, %0a, or raw CRLF bytes.
- SMTP relay logs showing malformed Content-Type parameters sourced from Roundcube accounts.
Detection Strategies
- Inspect Roundcube's sendmail.log and MTA submission logs for messages with abnormal Content-Type headers or injected Content-Disposition fields.
- Review HTTP request bodies to the compose endpoint for data: URIs whose MIME type field contains non-token characters.
- Compare the running Roundcube version against 1.6.20 and 1.7.5 across all webmail hosts.
Monitoring Recommendations
- Enable verbose SMTP submission logging on the MTA that fronts Roundcube to capture full MIME structure anomalies.
- Alert on repeated failed message deliveries or quarantine events referencing malformed MIME parts from authenticated webmail users.
- Track authenticated session activity for users sending large numbers of HTML messages with inline image data URIs.
How to Mitigate CVE-2026-114716
Immediate Actions Required
- Upgrade Roundcube Webmail to 1.6.20 or 1.7.5 without delay.
- Apply distribution-provided security updates where Roundcube is installed from packages, including those tracked in Debian Bug Report #1151279.
- Audit webmail accounts for suspicious draft or sent messages containing crafted data: URIs.
Patch Information
The fix is published in Roundcube releases 1.6.20 and 1.7.5 under the commit titled "Fix mail header injection via media type in an inline image data URI." The patch adds validation through rcube_mime::is_mimetype_valid() so any CRLF-tainted media type is replaced by a value derived from the image bytes. Refer to the Roundcube Security Updates Announcement for release notes and upgrade instructions.
Workarounds
- Restrict Roundcube access to trusted networks or require multi-factor authentication to limit the pool of accounts that could abuse the flaw.
- Disable HTML composition for untrusted user populations where feasible to prevent inline image attachment flows.
- Deploy an outbound mail content filter that rejects messages with malformed MIME headers until patching is complete.
# Example: verify installed Roundcube version on a Debian/Ubuntu host
dpkg-query -W -f='${Package} ${Version}\n' roundcube-core
# Example: identify Roundcube installations by release file
find /var/www -name 'index.php' -path '*/roundcube*' -exec \
grep -H 'RCMAIL_VERSION' {} \;
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

