CVE-2026-76894

A framed CKAN MCP table can display fabricated data

onData · CKAN MCP Server DataStore table UI

First documented disclosure · Review checkpoint

Summary

An untrusted frame can inject rows or imitate the host’s response during table initialization.

Severity and exploitation evidence

Low · 3.1/10 · CVSS 3.1. Severity source.

Exploitation not established: The researchers provide a controlled frame-spoofing example. No victim compromise is documented.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The embedded table consumes window messages without checking their origin and sends messages to its parent with a wildcard destination.

Conditions required

A frame can reach the embedded table, or an untrusted parent embeds it. A host that tightly controls embedding limits exposure.

Potential impact

Displayed fields, records and links can be forged. An untrusted parent may receive protocol messages. The renderer escapes the supplied values.

Affected versions and configurations

Versions before 0.4.112 that use the affected embedded UI.

Fix and mitigation status

0.4.112 learns an origin from the first matching handshake reply. After initialization, it filters data messages and targets outgoing messages to that origin.

Practical next steps

Update the server and restrict which frames can interact during startup. Outbound messages initially use a wildcard, and pending-response IDs are handled before origin filtering.

Operating lesson

Authenticate both directions of a webview’s message channel before trusting tool output.

Evidence and limits

The advisory, patch and release are dated July 9. The November 3 CVE covers this table UI issue only. Display spoofing was demonstrated without script execution.

Sources