/

CVE-2026-74998

Severity:

HIGH RISK

(7.2)

Cross-Site Scripting (CWE-79) in Roundcube CSS Proxy

CVE-2026-74998

Analyzed

Vulnerability type

Category:

Cross-Site Scripting

CWE:

CWE-79

Impact:

Client-Side Code Execution and Information Disclosure

Severity:

HIGH RISK

(7.2)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Roundcube Webmail did not adequately validate responses returned by its CSS proxy, allowing content to be interpreted under an unintended MIME type. Under conditions where attacker-influenced content reaches the proxy and is processed by a victim's browser, this may enable cross-site scripting or information disclosure.

Affected context

The issue occurs in Roundcube Webmail deployments using the CSS proxy to retrieve or serve remote stylesheet content. It may become risky when proxy responses are not constrained to the expected CSS content type, MIME handling permits browser interpretation as another active content type, and an attacker can influence the content or resource referenced through the proxy.

Exploit preconditions

  • Attack surface: Roundcube's CSS proxy and the browser handling its responses
  • Required access: An attacker generally needs a way to influence content retrieved through the proxy or cause a victim to access an affected resource
  • Trigger point: Processing a CSS proxy response without sufficient content or MIME validation
  • Why exploitation is possible: Browser MIME sniffing may cause attacker-controlled content to be interpreted as executable or otherwise active content rather than as inert CSS

Why this happens

The underlying bug is insufficient validation of data returned by a proxy endpoint. A response expected to contain CSS was not reliably constrained by content validation and MIME handling, allowing a browser to infer or honor a more dangerous interpretation.

This maps to CWE-79: Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting'), because attacker-influenced content may reach a browser in a context where script-capable interpretation becomes possible. The insecure pattern is trusting a proxied response's declared or inferred type instead of enforcing an allowlisted content type and safe browser handling.

Potential impact

  • An attacker may execute script in a victim's Roundcube browser context
  • Session-related information or webmail data could be exposed
  • Malicious actions could be performed through the victim's authenticated session
  • Users may be redirected to or served unexpected active content

General mitigation guidance

Mitigation typically focuses on treating proxied content as untrusted and enforcing its intended representation before it reaches the browser.

Common defensive patterns include:

  • Allowing only approved CSS media types and rejecting mismatched responses
  • Disabling MIME sniffing with appropriate response-handling controls
  • Applying strict output encoding and content-security policies where compatible
  • Restricting proxy destinations and reducing the ability to retrieve attacker-controlled resources
  • Running webmail components with least privilege and minimizing session exposure
  • Separating untrusted proxied content from authenticated application origins where practical

How teams detect issues like this

Organizations commonly detect this class of issue through:

  • Static analysis and code review checks for proxy responses lacking content-type validation
  • Review of response-header handling, MIME sniffing behavior, and browser-facing endpoints
  • Dynamic tests using benign content with mismatched media types to verify rejection or safe rendering
  • Web application firewall or gateway policies that restrict unexpected content types from proxy endpoints
  • Monitoring for unusual CSS proxy destinations, response types, and script-capable content

Automated application-security testing can help identify reflected or stored script execution paths, but should be supplemented with manual review of browser interpretation behavior.

Sources

SourceLink
github.comgithub.com/roundcube/roundcubemail/commit/62d33c8a0dc3fd0dd03984220dc9709e8e0de43b
github.comgithub.com/roundcube/roundcubemail/commit/a2334990c02f7cb77b1a8d6fa97fcd9e1b5cd1fb
github.comgithub.com/roundcube/roundcubemail/releases/tag/1.6.18
github.comgithub.com/roundcube/roundcubemail/releases/tag/1.7.3
roundcube.netroundcube.net/news/2026/08/09/security-updates-1.6.18-and-1.7.3

Frequently asked

questions

Exploitability depends on deployment behavior and whether an attacker can influence content retrieved through the CSS proxy or induce a victim to access a crafted resource. It is not necessarily exposed in configurations that disable or tightly restrict the proxy.

The relevant requirement is access to the proxy's reachable attack surface and a victim who processes the response. Whether an attacker must authenticate depends on the Roundcube deployment's access controls and proxy configuration.

Enforce strict allowlisted content types for CSS proxy responses, reject mismatches, and prevent browsers from MIME-sniffing responses into active content.

Treat all proxied data as untrusted, validate response bodies and headers against the expected resource type, isolate untrusted content from authenticated origins, and test browser-facing proxy endpoints for MIME confusion and XSS behavior.

Find out what's already
exploitable in your codebase.