Description

The TranslatePress – Translate Multilingual sites with AI Translation plugin for WordPress is vulnerable to Stored Cross-Site Scripting via Comment Noise-Key Injection into HTML Parser in all versions up to, and including, 3.3.3 due to insufficient input sanitization and output escaping. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. Exploitation is possible because WordPress's comment KSES allowlist permits the payload structure — an anchor tag with href and title attributes alongside a code tag — causing the malicious comment to be stored verbatim in the database, where it is later processed by the vulnerable parser during page translation.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "7.2",
  "pubDate": "2026-08-28T05:16:46.087Z",
  "pubdate": "2026-08-28T05:16:46.087Z",
  "executiveSummary": "The TranslatePress – Translate Multilingual sites with AI Translation plugin for WordPress is vulnerable to a critical Stored Cross-Site Scripting (XSS) vulnerability in all versions up to and including 3.3.3. This security flaw arises from a combination of insufficient input sanitization, inadequate output escaping, and a specific parser behavior during the translation process. Unauthenticated attackers can exploit this vulnerability by injecting malicious payloads via public comment sections on affected sites.\nBecause the payload structure complies with the default WordPress comment KSES allowlist—which permits specific HTML elements such as anchor tags with href and title attributes alongside code tags—the malicious input successfully bypasses initial filters and is stored verbatim within the database. When a user or administrator subsequently visits the affected page, the plugin's vulnerable HTML parser processes this stored content, triggering the execution of arbitrary JavaScript. The execution of unauthorized scripts in the context of a victim's browser can lead to complete session hijacking, unauthorized administrative actions, and potential site compromise.",
  "technicalDetails": "The vulnerability lies within the interaction between the WordPress core comment filtering mechanism (KSES) and the HTML parser implemented by the TranslatePress plugin during the translation phase. When a comment is submitted by an unauthenticated user, WordPress subjects the input to the KSES allowlist. This allowlist natively permits certain HTML structures to allow basic formatting in comments, specifically allowing anchor tags containing 'href' and 'title' attributes, as well as code tags.\nAn attacker crafts a specific payload that perfectly aligns with these allowed HTML structures, embedding malicious scripts within the permitted attributes or tags. Because the payload matches the KSES rules, the WordPress database stores the comment verbatim, containing the hidden injection vector. The core issue manifests when the TranslatePress plugin retrieves this stored comment from the database to translate the page content.\nDuring the page rendering process, TranslatePress passes the HTML content, including the stored comment, through its internal translation parser. Due to a failure to properly sanitize and escape noise keys or specific attributes within the parser, the parser incorrectly processes the structured HTML. This processing failure reconstructs the HTML in a way that executes the injected JavaScript code.\nThe attack flow occurs as follows: First, an unauthenticated remote attacker submits a comment containing the crafted anchor and code tags to a post. Second, the comment bypasses WordPress KSES filtering and is written to the database. Third, when any subsequent visitor, including highly privileged administrators, requests the translated version of that page, the TranslatePress translation engine parses the database content. Fourth, the vulnerability in the parser triggers, executing the attacker's script in the victim's browser session. The impact is significant, as executing script in an administrator's session can allow attackers to create rogue admin accounts, install malicious plugins, or deface the website."
}