Description

Mattermost Plugins versions <=11.8 10.20.11 11.5.7.0 _The Mattermost GitLab plugin fails to verify channel permissions when processing API requests with a caller-supplied_ {{post_id}}_, and fails to validate the_ {{web_url}} _parameter against the configured GitLab instance, which allows an authenticated attacker to inject bot-authored messages containing arbitrary URLs into channels they do not have access to via the_ {{createIssue}} _and_ {{attachCommentToIssue}} _API endpoints._ Mattermost Advisory ID: MMSA-2026-00673

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "4.3",
  "pubDate": "2026-08-17T15:16:54.420Z",
  "pubdate": "2026-08-17T15:16:54.420Z",
  "executiveSummary": "An authorization and validation vulnerability exists in the Mattermost GitLab plugin within versions <=11.8, 10.20.11, and 11.5.7.0. The vulnerability allows an authenticated attacker to inject bot-authored messages containing arbitrary URLs into restricted channels.\nThe core issues stem from a failure to verify channel permissions when processing API requests utilizing a caller-supplied post_id, alongside an insufficient validation mechanism for the web_url parameter against the configured GitLab instance.\nAffected systems include Mattermost deployments utilizing vulnerable versions of the GitLab plugin. The risk implications involve unauthorized data injection, potential phishing vectors, and the compromise of channel message integrity across restricted communication perimeters.\nAn attacker must possess valid authentication within the Mattermost instance to leverage the createIssue and attachCommentToIssue API endpoints. No specific elevated privileges are required beyond standard user authentication to exploit the missing channel access controls.\nExploitation allows unauthorized actors to bypass security boundaries, injecting arbitrary crafted content and external links into channels for which they lack membership or explicit access permissions.",
  "technicalDetails": "The vulnerability resides in the Mattermost GitLab plugin API endpoints, specifically within the handling logic for the createIssue and attachCommentToIssue functions. The vulnerable components fail to execute adequate authorization checks on the caller-supplied post_id parameter, neglecting to validate whether the authenticated caller possesses legitimate access or permissions to the target channel where the operation is initiated.\nFurthermore, the plugin fails to enforce proper input validation on the web_url parameter. The system does not validate the provided URL against the securely configured GitLab instance whitelist or domain parameters, enabling the acceptance of arbitrary external URLs.\nThe attack flow proceeds as follows: First, an authenticated attacker crafts a malicious API request targeting either the createIssue or attachCommentToIssue endpoints exposed by the GitLab plugin. Within the request payload, the attacker supplies a manipulated post_id to bypass intended channel permission boundaries, alongside a malicious web_url parameter pointing to an arbitrary, attacker-controlled destination.\nUpon receiving the request, the vulnerable plugin fails to verify channel membership or permissions associated with the caller and the target post_id. It subsequently processes the input without validating the web_url against the legitimate GitLab instance configuration.\nThe plugin then uses its bot credentials to generate and dispatch a bot-authored message containing the injected arbitrary URLs directly into the unauthorized channel.\nNetwork exposure is restricted to authenticated users interacting with the Mattermost API endpoints exposed by the GitLab plugin. Authentication is required to interact with the API, but privilege requirements are limited to standard authenticated users who can invoke the affected createIssue and attachCommentToIssue functions.\nPost-exploitation impact includes the dissemination of arbitrary URLs, potential social engineering attacks, phishing, and unauthorized message injection within channels that the attacker is otherwise forbidden to access."
}