CVE-2026-11436: Mage AI XSS Vulnerability

CVE-2026-11436 Overview

CVE-2026-11436 is a reflected cross-site scripting (XSS) vulnerability in Mage AI through version 0.9.79. The flaw resides in the useMutation function within mage_ai/frontend/components/Sessions/SignForm/index.tsx, part of the Sign-in Flow component. An attacker can manipulate the query.redirect_url parameter to inject malicious script content that executes in the victim's browser. The exploit is publicly available, and the vendor did not respond to disclosure attempts. The vulnerability is tracked under CWE-79 and is exploitable remotely with user interaction.

Critical Impact
Successful exploitation enables attackers to execute arbitrary JavaScript in the context of a victim's authenticated Mage AI session, enabling session token theft, credential capture, and phishing attacks targeting data engineers using the platform.

Affected Products

Mage AI versions up to and including 0.9.79
Component: Sign-in Flow (mage_ai/frontend/components/Sessions/SignForm/index.tsx)
Function: useMutation processing the query.redirect_url argument

Discovery Timeline

2026-06-06 - CVE-2026-11436 published to NVD
2026-06-08 - Last updated in NVD database

Technical Details for CVE-2026-11436

Vulnerability Analysis

The vulnerability is a client-side reflected XSS flaw in the Mage AI authentication interface. The useMutation hook within the SignForm component consumes the redirect_url query string parameter without proper sanitization or output encoding. When a user visits a crafted sign-in URL, attacker-controlled content in redirect_url is reflected into the page and executed by the browser.

Exploitation requires user interaction, typically clicking a crafted link. The attack vector is the network, and no privileges are required to deliver the payload. Impact is limited to integrity of the rendered page, with no direct effect on backend confidentiality or availability. A public proof of concept is hosted on GitHub Gist.

Root Cause

The root cause is improper neutralization of input during web page generation [CWE-79]. The redirect_url value flows from the URL query string into the React component state and downstream operations within useMutation without escaping or validation against an allowlist of trusted redirect targets. Mage AI's frontend trusts the parameter as a navigation target rather than treating it as untrusted input.

Attack Vector

An attacker crafts a URL pointing to a legitimate Mage AI deployment's sign-in page with a malicious redirect_url query parameter containing a JavaScript payload. The attacker delivers the link through phishing, social media, or a compromised site. When a Mage AI user clicks the link, the payload executes in the browser within the application's origin, exposing session cookies, stored tokens, and sensitive pipeline metadata.

No verified exploit code is reproduced here. Refer to the public proof of concept and the VulDB entry for CVE-2026-11436 for technical specifics.

Detection Methods for CVE-2026-11436

Indicators of Compromise

Web server or proxy logs containing requests to Mage AI sign-in endpoints with a redirect_url query parameter containing <script>, javascript:, onerror=, onload=, or URL-encoded equivalents (%3Cscript%3E, %6A%61%76%61%73%63%72%69%70%74).
Inbound HTTP referrers pointing to untrusted external domains immediately preceding sign-in page loads.
Browser console errors or unexpected outbound requests from authenticated Mage AI sessions to attacker-controlled domains.

Detection Strategies

Deploy a web application firewall (WAF) rule that inspects the redirect_url parameter on all paths under the Mage AI sign-in route and blocks payloads matching common XSS signatures.
Enable Content Security Policy (CSP) reporting on the Mage AI deployment and forward violation reports to a centralized log store for analysis.
Hunt across HTTP access logs for sign-in URLs where redirect_url contains characters outside an expected allowlist such as scheme, host, and path.

Monitoring Recommendations

Alert on user accounts that authenticate and then immediately issue unusual API calls or token refreshes from new IP addresses, which may indicate a hijacked session.
Monitor frontend asset integrity and CSP violation telemetry for repeated reports tied to the sign-in route.
Correlate phishing-themed inbound emails referencing Mage AI URLs with subsequent authentication events from targeted users.

How to Mitigate CVE-2026-11436

Immediate Actions Required

Restrict public network exposure of Mage AI sign-in pages by placing the application behind a VPN or authenticated reverse proxy until a patched release is available.
Deploy WAF rules that strip or reject redirect_url query parameters not matching an internal allowlist of trusted paths.
Notify Mage AI users of the active public exploit and instruct them not to click sign-in links received from external sources.

Patch Information

No official vendor patch has been published. According to the CVE record, the vendor was contacted early about this disclosure but did not respond. Operators running Mage AI 0.9.79 or earlier should monitor the Mage AI GitHub repository for fixed releases and apply updates immediately upon availability. In the interim, consider forking and applying a local fix that validates redirect_url against an allowlist of relative paths and rejects absolute URLs and javascript: schemes.

Workarounds

Implement a strict Content Security Policy on the Mage AI deployment that disallows inline scripts and restricts script sources to the application's own origin.
Configure the reverse proxy to drop or sanitize the redirect_url query parameter on requests to the sign-in route.
Rotate session secrets and force re-authentication for all users after deploying mitigations to invalidate any tokens that may have been stolen via this XSS.
Train users to inspect sign-in URLs and report suspicious links to the security team.

bash

# Example NGINX configuration to strip the redirect_url parameter on sign-in requests
location /sign-in {
    if ($arg_redirect_url != "") {
        return 302 /sign-in;
    }
    proxy_pass http://mage_ai_backend;
}