CVE-2025-43849

9.8 CRITICAL

📋 TL;DR

This vulnerability allows remote attackers to execute arbitrary code on systems running Retrieval-based-Voice-Conversion-WebUI by exploiting unsafe deserialization in the model loading process. Attackers can achieve remote code execution by providing malicious model files or paths. Users of versions 2.2.231006 and earlier are affected.

💻 Affected Systems

Products:
  • Retrieval-based-Voice-Conversion-WebUI
Versions: Versions 2.2.231006 and prior
Operating Systems: All platforms running Python and PyTorch
Default Config Vulnerable: ⚠️ Yes
Notes: Any deployment using the vulnerable merge functionality with user-controlled model paths is affected.

📦 What is this software?

⚠️ Risk & Real-World Impact

🔴

Worst Case

Full system compromise with attacker gaining complete control over the server, allowing data theft, lateral movement, and persistent backdoor installation.

🟠

Likely Case

Remote code execution leading to service disruption, cryptocurrency mining, or data exfiltration from the vulnerable system.

🟢

If Mitigated

Limited impact if system is isolated with minimal privileges, though RCE could still affect the application's data and functionality.

🌐 Internet-Facing: HIGH
🏢 Internal Only: MEDIUM

🎯 Exploit Status

Public PoC: ✅ No
Weaponized: LIKELY
Unauthenticated Exploit: ⚠️ Yes
Complexity: LOW

The vulnerability is straightforward to exploit as it involves passing malicious input to torch.load() without proper validation.

🛠️ Fix & Mitigation

✅ Official Fix

Patch Version: None

Vendor Advisory: https://securitylab.github.com/advisories/GHSL-2025-012_GHSL-2025-022_Retrieval-based-Voice-Conversion-WebUI/

Restart Required: Yes

Instructions:

No official patch exists. Monitor the GitHub repository for updates and apply when available.

🔧 Temporary Workarounds

Input Validation and Sanitization

all

Implement strict validation of user-provided model paths to prevent directory traversal and ensure only trusted model files are loaded.

Network Isolation

all

Restrict network access to the application to trusted users only using firewall rules.

🧯 If You Can't Patch

  • Disable the vulnerable merge functionality entirely if not required.
  • Run the application in a sandboxed environment with minimal privileges and no network access.

🔍 How to Verify

Check if Vulnerable:

Check if using Retrieval-based-Voice-Conversion-WebUI version 2.2.231006 or earlier by examining the version in the application or configuration files.

Check Version:

Check the version in the application's source code or configuration files (no standard command).

Verify Fix Applied:

Verify that the application version is later than 2.2.231006 once a patch is released.

📡 Detection & Monitoring

Log Indicators:

  • Unusual model file paths in application logs
  • Errors from torch.load() with suspicious inputs
  • Unexpected process execution from the application

Network Indicators:

  • Unexpected outbound connections from the application server
  • Suspicious file uploads to model endpoints

SIEM Query:

Search for process execution events from the RVC-WebUI application or related Python processes with unusual command-line arguments.

🔗 References

📤 Share & Export