Applio unsafe deserialization vulnerability in model_information.py
GitHub Security Lab identified unsafe deserialization vulnerabilities in Applio 3.2.7, particularly in files using torch.load() without the weights_only parameter. The maintainers fixed the issue in February 2025 by adding the weights_only=True parameter to restrict deserialization to tensor data only.
What happened
Applio version 3.2.7 was found to contain unsafe deserialization vulnerabilities in model_information.py and other files that use torch.load() without the weights_only parameter.
GitHub Security Lab reported the vulnerability via coordinated disclosure on 2024-12-11, with fixes applied on 2025-02-01 to add weights_only=True to all torch.load() calls.
Reported impact
Unsafe deserialization could be exploited to achieve remote code execution on the Applio server when processing untrusted model files.
Index analysis
The vulnerability exists in torch.load() calls across multiple files including model_information.py, infer.py, FCPE.py, RMVPE.py, and data_utils.py that deserialize model weights without restricting deserialization to safe tensor data.
Governance controls
The maintainer applied fixes on 2025-02-01 by adding the weights_only=True parameter to torch.load() calls, which restricts deserialization to tensor data only and prevents arbitrary code execution.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.