CVE-2026-6592: ComfyUI XSS Vulnerability

CVE-2026-6592 Overview

A cross-site scripting (XSS) vulnerability has been discovered in ComfyUI, a popular open-source AI image generation interface. The vulnerability exists in the getuserdata function within the app/user_manager.py file of the userdata endpoint. This flaw allows remote attackers to inject malicious scripts through improper handling of user-supplied input, potentially compromising user sessions and sensitive data within the application context.

Critical Impact

Remote attackers can execute arbitrary JavaScript code in the context of authenticated user sessions, potentially leading to session hijacking, credential theft, or unauthorized actions within the ComfyUI interface.

Affected Products

• ComfyUI versions up to and including 0.13.0

Discovery Timeline

• April 20, 2026 - CVE-2026-6592 published to NVD
• April 22, 2026 - Last updated in NVD database

Technical Details for CVE-2026-6592

Vulnerability Analysis

This vulnerability is classified as CWE-79 (Improper Neutralization of Input During Web Page Generation), commonly known as Cross-Site Scripting (XSS). The flaw resides in the getuserdata function within app/user_manager.py, which serves as part of the userdata endpoint in ComfyUI. The function fails to properly sanitize or encode user-controlled input before rendering it in the response, allowing attackers to inject malicious script content.

The vulnerability requires network access and some level of authentication (low privileges), along with user interaction to trigger the malicious payload. When successfully exploited, attackers can execute arbitrary JavaScript code within the victim's browser session, potentially accessing session tokens, cookies, and other sensitive information stored in the browser context.

Root Cause

The root cause of this vulnerability is insufficient input validation and output encoding in the getuserdata function. When user-supplied data is processed by the userdata endpoint, the application fails to properly sanitize special characters that could be interpreted as HTML or JavaScript by the browser. This allows attackers to craft malicious input that, when reflected back to other users, executes arbitrary scripts in their browser context.

Attack Vector

The attack is network-based and can be executed remotely against vulnerable ComfyUI installations. An attacker with low-level privileges can craft a specially formatted request to the userdata endpoint containing malicious JavaScript payloads. When another user interacts with the manipulated data or views a page containing the unsanitized output, the injected script executes in their browser session.

The exploitation mechanism involves sending crafted input through the userdata API endpoint that bypasses input validation. The malicious payload is then stored or reflected through the getuserdata function, and when rendered in a victim's browser, the JavaScript code executes with the same origin permissions as the ComfyUI application.

A proof-of-concept demonstrating this vulnerability has been publicly disclosed. For technical details, refer to the GitHub Gist PoC.

Detection Methods for CVE-2026-6592

Indicators of Compromise

• Unusual JavaScript or HTML tags appearing in userdata API request parameters
• Unexpected outbound network connections originating from user browser sessions
• Log entries showing requests to /userdata endpoints containing script tags, event handlers, or encoded JavaScript
• User reports of unexpected behavior or pop-ups when accessing ComfyUI interfaces

Detection Strategies

• Implement web application firewall (WAF) rules to detect and block common XSS payload patterns in requests to the userdata endpoint
• Monitor server access logs for requests containing suspicious characters or encoded script content targeting app/user_manager.py routes
• Deploy client-side Content Security Policy (CSP) violation monitoring to detect attempted script injection
• Utilize SentinelOne Singularity platform to detect anomalous process behavior and network connections from browser processes

Monitoring Recommendations

• Enable verbose logging for the ComfyUI userdata endpoint and review logs regularly for injection attempts
• Configure alerting for requests containing common XSS indicators such as <script>, javascript:, onerror=, and similar patterns
• Implement real-time monitoring of outbound connections from systems running ComfyUI to detect potential data exfiltration

How to Mitigate CVE-2026-6592

Immediate Actions Required

• Restrict access to ComfyUI installations to trusted networks only until a patch is available
• Implement a web application firewall with XSS protection rules in front of ComfyUI deployments
• Review and audit user data stored in the application for any signs of injected malicious content
• Educate users about the risks of clicking on untrusted links or interacting with suspicious content within the application

Patch Information

As of the last CVE update, the vendor (ComfyUI) was contacted about this disclosure but did not respond. No official patch is currently available. Users should monitor the official ComfyUI repository and VulDB advisory for updates regarding security fixes.

Workarounds

• Deploy a reverse proxy or WAF with XSS filtering capabilities to sanitize input before it reaches the ComfyUI application
• Implement Content Security Policy (CSP) headers to restrict script execution sources and mitigate the impact of successful XSS attacks
• Restrict network access to ComfyUI instances using firewall rules, limiting exposure to trusted IP ranges only
• Consider disabling or restricting access to the userdata endpoint if not critical to operations

# Example nginx configuration to add basic XSS protection headers
server {
    listen 80;
    server_name comfyui.example.com;
    
    # Add security headers
    add_header X-Content-Type-Options "nosniff" always;
    add_header X-XSS-Protection "1; mode=block" always;
    add_header Content-Security-Policy "default-src 'self'; script-src 'self'; style-src 'self' 'unsafe-inline';" always;
    
    location / {
        proxy_pass http://localhost:8188;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
    }
}