Whispering Under the Eaves: Protecting User Privacy Against Commercial and LLM-powered Automatic Speech Recognition Systems
Year: 2025
Type: conference-paper
Abstract: Here are the materials related to AudioShield, including: source_code.zip: The source code of AudioShield. datasets.zip: The training and testing datasets, processed in pickle format and ready for use. The 500 samples from Librispeech are used for training, and 2000 samples from VCTK are used for testing. pretrained_models.zip: The pre-trained models required for training and testing, including the locally trained target ASR model, DeepSpeech, and the Autoencoder model VITS used by AudioShield. Please unzip the files in datasets.zip into the source_code/datasets folder, and unzip the files in pretrained_models.zip into the source_code/pretrained folder. For detailed installation and usage instructions, please refer to the README file in the source_code folder.
Author: Jin, Weifei
Institution: Beijing University of Posts and Telecommunications
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6949285777
FWCI:
Cites:
Cited by:
Related to: