Description

Amundsen frontend through 4.3.0 renders table, dashboard, and feature descriptions with dangerouslySetInnerHTML without HTML sanitization in ResourceListItem components. Attackers can inject malicious markup like img elements with onerror handlers into descriptions via the metadata service or Elasticsearch, executing JavaScript in every user's browser that views search results.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "7.6",
  "pubDate": "2026-09-13T11:17:01.780Z",
  "pubdate": "2026-09-13T11:17:01.780Z",
  "executiveSummary": "This vulnerability is a Stored Cross-Site Scripting (XSS) flaw identified within the Amundsen frontend, affecting versions up to 4.3.0. The vulnerability originates from the insecure utilization of React's dangerouslySetInnerHTML property to render dynamic content, specifically table, dashboard, and feature descriptions, without prior HTML sanitization.\nBy injecting malicious markup into the metadata service or Elasticsearch index, an attacker can ensure that arbitrary JavaScript is executed within the browser context of any user viewing the affected search results. The impact is significant, as it facilitates unauthorized access to sensitive user sessions, credential theft, or the execution of unauthorized actions on behalf of the victim. Given that Amundsen functions as a data discovery and metadata engine, successful exploitation could lead to broader information disclosure. Exploitation requires the attacker to have the capability to influence metadata content, typically via the metadata ingestion pipeline or direct interaction with the underlying Elasticsearch instance, though no specific authentication is required if the ingestion process is reachable or improperly secured.",
  "technicalDetails": "The root cause of this vulnerability lies in the improper handling of user-controlled input within the ResourceListItem component of the Amundsen frontend. The application fetches metadata strings—specifically table, dashboard, and feature descriptions—from the metadata service or Elasticsearch backend and renders them directly into the Document Object Model (DOM) using the dangerouslySetInnerHTML React property. Because this operation bypasses React's default XSS protection, which typically escapes content, any HTML tags or script fragments embedded within the metadata are parsed and executed by the browser.\nThe attack flow begins with the injection of a malicious payload into the metadata storage layer. An attacker leverages an entry point—such as an ingestion script, an API interaction, or a direct write to Elasticsearch—to submit a crafted string containing HTML elements. A common attack vector involves embedding an <img> tag with a malformed source attribute and an 'onerror' event handler: <img src=x onerror='alert(document.cookie)'>. When the browser attempts to render the description, the image fails to load, triggering the execution of the JavaScript payload within the security context of the user's current session.\nThe vulnerable component is identified within the frontend rendering logic, specifically where ResourceListItem components process metadata descriptions. Since Amundsen aggregates metadata from various sources to provide a searchable discovery experience, the attack surface is broad. The vulnerability persists across all versions up to 4.3.0. There are no client-side restrictions preventing the rendering of these untrusted strings, and the absence of a sanitization library (such as DOMPurify) on the frontend allows for the successful injection of complex scripts.\nPost-exploitation, the impact is severe. Since the script executes within the victim's session, an attacker can perform actions that include hijacking session tokens, stealing CSRF tokens, exfiltrating sensitive metadata, or performing unauthorized API requests to the Amundsen backend. Because these descriptions are often displayed prominently in search results, the exploit is highly effective and can be deployed at scale, potentially affecting any user who performs a search query that matches the poisoned metadata entry."
}