A tool to evaluate the distinctiveness of voiceprint via Content-based PROLE Score.
This is the source code of the tool to calculate the PROLE Score on the Web site. We define the PROLE Score as a metric to calculate the distinctiveness of voiceprint and it depends on the model and speech content used to generate the voiceprint. It is well known, voiceprints with high distinctiveness have higher security, so that the PROLE Score can be used to evaluate the security of voiceprint.
You can learn about the algorithm of the tool from the paper "OK, Siri" or "Hey, Google": Evaluating Voiceprint Distinctiveness via Content-based PROLE Score in the USENIX Security 2022. And if you use the tool or the method in your work, please add the following citation.
@inproceedings{He2022ok,
title={"OK, Siri" or "Hey, Google": Evaluating Voiceprint Distinctiveness via Content-based *PROLE Score*"},
author={He, Ruiwen and Ji, Xiaoyu and Li, Xinfeng and Cheng, Yushi and Xu, Wenyuan},
booktitle={USENIX Security Symposium},
year={2022}
}
If you want to add this tool to your website, you need to follow these steps:
-
Add libraries as requirements
pip install -r requirements.txt
-
Running
python app.py
Besides, if you just want to use the tool to evaluate your speech content used to generate the voiceprint, for example by selecting some appropriate wake-up words, you can leverage the tool in our website. All you need to do is enter the speech content and select a voiceprint model, and you will get the PROLE Score and suggestions for improving the phrases.