GHSA-9Q82-XGWF-VJ6H

Apollo Server: Browser bug allows for bypass of XS-Search (read-only Cross-Site Request Forgery) prevention

app.use((req, res, next) => {
  for (let i = 0; i < req.rawHeaders.length - 1; i += 2) {
    if (
      req.rawHeaders[i].toLowerCase() === 'content-type' &&
      req.rawHeaders[i + 1].includes('message/')
    ) {
      return res.status(415).json({ error: 'Content-Type not allowed' });
    }
  }
  next();
});

In a Cross-Site Request Forgery attack, untrusted web content causes browsers to send authenticated requests to web servers which use cookies for authentication. While the web content is prevented from reading the request's response due to the Cross-Origin Request Sharing (CORS) protocol, an attacker may be able to cause side effects in the server ("CSRF" attack), or learn something about the response via timing analysis ("XS-Search" attack).

Apollo Server has a built-in feature which prevents CSRF and XS-Search attacks: it refuses to process GraphQL requests that could possibly have been sent by a spec-compliant web browser without a protective "preflight" step. See Apollo Server's docs for more details on CORS, CSRF attacks, and Apollo Server's CSRF prevention feature.

This feature is fully effective against attacks carried out against users of spec-compliant browsers. Unfortunately, a major browser introduced a bug in 2025 which meant in certain cases, it failed to follow the CORS spec. The browser's maintainers have already committed to fixing the bug and making the browser spec-compliant again.

Even with this bug, Apollo Server's CSRF prevention feature blocks "side effect" CSRF attacks: Apollo Server will still correctly refuse to execute mutations in requests that were not preflighted. However, some specially crafted authenticated GraphQL queries can be issued across origins without preflight in buggy versions of this browser, allowing for XS-Search attacks: an attacker can analyze response times to learn facts about the responses to requests such as whether fields return null or approximately how many list entries are returned from fields.

GraphQL servers are only vulnerable if they rely on cookies (or HTTP Basic Auth) for authentication.

Kodem intelligence

Already deployed Kodem?

The vulnerability is patched in @apollo/server v5.5.0. This release contains a single change: GraphQL requests sent in HTTP GET requests which contain a Content-Type header naming a type other than application/json are rejected. (GET requests with no Content-Type are allowed.) This change prevents XS-Search attacks even in browsers which are non-compliant in ways similar to this browser.

There are no known cases where GraphQL apps depend on the ability of clients to send non-empty Content-Type headers with GET requests other than application/json, so this change has not been made configurable; if this change breaks a use case, file an issue and more configurability can be added.

Apollo is not currently providing a patch for previous major versions of Apollo Server, which are all end-of-life.

1. What is GHSA-9Q82-XGWF-VJ6H? GHSA-9Q82-XGWF-VJ6H is a medium-severity security vulnerability in @apollo/server (npm), affecting versions < 5.5.0. It is fixed in 5.5.0.
2. Which packages are affected by GHSA-9Q82-XGWF-VJ6H?
   • @apollo/server (npm) (versions < 5.5.0)
   • apollo-server-core (npm) (versions <= 3.13.0)
3. Is there a fix for GHSA-9Q82-XGWF-VJ6H? Yes. GHSA-9Q82-XGWF-VJ6H is fixed in 5.5.0. Upgrade to this version or later.
4. Is GHSA-9Q82-XGWF-VJ6H exploitable, and should I be worried? Whether GHSA-9Q82-XGWF-VJ6H is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
5. What actually determines whether GHSA-9Q82-XGWF-VJ6H is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
6. How do I fix GHSA-9Q82-XGWF-VJ6H? Upgrade @apollo/server to 5.5.0 or later.

Other vulnerabilities in @apollo/server