CVE-2025-43852 Information

Description

Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Versions 2.2.231006 and prior are vulnerable to unsafe deserialization. The model_choose variable takes user input (e.g. a path to a model) and passes it to the uvr function in vr.py. In uvr  if model_name contains the string \DeEcho\ a new instance of AudioPreDeEcho class is created with the model_path attribute containing the aforementioned user input. In the AudioPreDeEcho class the user input is used to load the model on that path with torch.load which can lead to unsafe deserialization and remote code execution. As of time of publication no known patches exist.

Reference

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/7ef19867780cf703841ebafb565a4e47d1ea86ff/infer/modules/uvr5/vr.py#L214 https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/9f2f0559e6932c10c48642d404e7d2e771d9db43/infer-web.py#L1073 https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/9f2f0559e6932c10c48642d404e7d2e771d9db43/infer-web.py#L1098 https://securitylab.github.com/advisories/GHSL-2025-012_GHSL-2025-022_Retrieval-based-Voice-Conversion-WebUI/