CVE-2026-23516
📋 TL;DR
This cross-site scripting (XSS) vulnerability in CVAT allows attackers to execute arbitrary JavaScript in victims' browser sessions by creating malicious labels or SVG images. Users of CVAT versions 2.2.0 through 2.54.0 are affected when they interact with attacker-controlled content. Successful exploitation gives attackers temporary access to all CVAT resources the victim can access.
💻 Affected Systems
- CVAT (Computer Vision Annotation Tool)
📦 What is this software?
⚠️ Risk & Real-World Impact
Worst Case
Complete compromise of CVAT instance with attacker gaining administrative privileges, data exfiltration, and potential lateral movement to connected systems.
Likely Case
Unauthorized access to sensitive annotation data, project manipulation, and privilege escalation within the CVAT environment.
If Mitigated
Limited impact with proper input validation and output encoding preventing successful exploitation.
🎯 Exploit Status
Requires authenticated attacker to create malicious content and social engineering to lure victims.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 2.55.0
Vendor Advisory: https://github.com/cvat-ai/cvat/security/advisories/GHSA-3m7p-wx65-c7mp
Restart Required: Yes
Instructions:
1. Backup CVAT data and configuration. 2. Update CVAT to version 2.55.0 using your deployment method (Docker, Kubernetes, etc.). 3. Restart all CVAT services. 4. Verify the update completed successfully.
🔧 Temporary Workarounds
Input Validation Enhancement
allImplement additional input validation for label names and SVG uploads
# Requires code modification - implement proper sanitization of user inputs
Content Security Policy
allImplement strict CSP headers to limit script execution
# Add to web server configuration: Content-Security-Policy: script-src 'self'
🧯 If You Can't Patch
- Restrict user permissions to prevent label creation by untrusted users
- Implement web application firewall rules to detect and block XSS payloads
🔍 How to Verify
Check if Vulnerable:
Check CVAT version via web interface or API endpoint /api/v1/server/aboutCheck Version:
curl -s http://cvat-server/api/v1/server/about | grep versionVerify Fix Applied:
Confirm version is 2.55.0 or higher and test XSS payloads are properly sanitized📡 Detection & Monitoring
Log Indicators:
- Unusual label creation patterns
- Multiple failed SVG upload attempts
- JavaScript execution in unexpected contexts
Network Indicators:
- Suspicious POST requests to label creation endpoints
- Unusual SVG file uploads
SIEM Query:
source="cvat" AND (event="label_creation" OR event="svg_upload") AND payload CONTAINS "<script>"