CVE-2024-7990
📋 TL;DR
A stored XSS vulnerability in open-webui version 0.3.8 allows attackers to inject malicious scripts via the model description field. When executed, these scripts can compromise user sessions, steal credentials, or perform unauthorized actions. All users of affected versions are vulnerable, including administrators.
💻 Affected Systems
- open-webui/open-webui
📦 What is this software?
Open Webui by Openwebui
⚠️ Risk & Real-World Impact
Worst Case
Administrator account takeover leading to full system compromise, data exfiltration, and arbitrary code execution on the server.
Likely Case
Session hijacking, credential theft, and unauthorized actions performed by authenticated users.
If Mitigated
Limited to client-side impact if proper input validation and output encoding are implemented.
🎯 Exploit Status
Exploitation requires access to create or modify models via the API endpoint.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: Version after 0.3.8
Vendor Advisory: https://huntr.com/bounties/2256e336-0f67-449e-a82d-7fc57081a21c
Restart Required: No
Instructions:
1. Update open-webui to the latest version. 2. Verify the fix by checking that model description inputs are properly sanitized.
🔧 Temporary Workarounds
Input Validation Filter
allImplement server-side input validation to sanitize HTML/JavaScript in model description fields.
🧯 If You Can't Patch
- Disable or restrict access to the /api/v1/models/add endpoint.
- Implement a web application firewall (WAF) with XSS protection rules.
🔍 How to Verify
Check if Vulnerable:
Test if unsanitized HTML/JavaScript in model description fields executes when rendered in chat.Check Version:
Check the open-webui version in the application interface or configuration files.Verify Fix Applied:
Confirm that script tags and JavaScript in model descriptions are properly escaped or removed.📡 Detection & Monitoring
Log Indicators:
- Unusual model creation/modification requests with script tags or JavaScript payloads.
Network Indicators:
- HTTP POST requests to /api/v1/models/add containing suspicious payloads.
SIEM Query:
source="webui_logs" AND (uri="/api/v1/models/add" AND (payload CONTAINS "<script>" OR payload CONTAINS "javascript:"))