The VoicePrivacy 2020 Challenge: Results and findings

• Published in: Computer speech & language Vol. 74; p. 101362
• Main Authors: Tomashenko, Natalia, Wang, Xin, Vincent, Emmanuel, Patino, Jose, Srivastava, Brij Mohan Lal, Noé, Paul-Gauthier, Nautsch, Andreas, Evans, Nicholas, Yamagishi, Junichi, O’Brien, Benjamin, Chanclu, Anaïs, Bonastre, Jean-François, Todisco, Massimiliano, Maouche, Mohamed
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.07.2022; Elsevier
• Subjects: Anonymization; Artificial Intelligence; Attack model; Automatic speech recognition; Computer Science; Metrics; Privacy; Speaker verification; Speech synthesis; Utility; Voice conversion; Privacy; Anonymization; Metrics; Utility; Voice conversion; Speech synthesis; Speaker verification; Attack model; Automatic speech recognition; speaker verification; anonymization; speech synthesis; attack model; utility; privacy; metrics; automatic speech recognition; voice conversion
• ISSN: 0885-2308; 1095-8363
• DOI: 10.1016/j.csl.2022.101362

This paper presents the results and analyses stemming from the first VoicePrivacy 2020 Challenge which focuses on developing anonymization solutions for speech technology. We provide a systematic overview of the challenge design with an analysis of submitted systems and evaluation results. In particular, we describe the voice anonymization task and datasets used for system development and evaluation. Also, we present different attack models and the associated objective and subjective evaluation metrics. We introduce two anonymization baselines and provide a summary description of the anonymization systems developed by the challenge participants. We report objective and subjective evaluation results for baseline and submitted systems. In addition, we present experimental results for alternative privacy metrics and attack models developed as a part of the post-evaluation analysis. Finally, we summarize our insights and observations that will influence the design of the next VoicePrivacy challenge edition and some directions for future voice anonymization research. •Systematic overview of the 1st VoicePrivacy Challenge: anonymization solutions for speech technology•The anonymization task is formulated as a game between an attacker and user; 3 attack models were investigated•18 anonymization systems; 2 classes of methods: using x-vectors with speech synthesis and signal-processing based transformations•Objective and subjective evaluation results are reported in terms of privacy and utility metrics•Insights for the next VoicePrivacy challenge and directions for future voice anonymization research