Inference Attacks for X-Vector Speaker Anonymization
We revisit the privacy-utility tradeoff of x-vector speaker anonymization. Existing approaches quantify privacy through training complex speaker verification or identification models that are later used as attacks. Instead, we propose a novel inference attack for de-anonymization. Our attack is simple and ML-free yet we show experimentally that it outperforms existing approaches.
Authors
- Vincent Bindschaedler (ORCID: https://orcid.org/0000-0002-3066-7354)
- Luke A. Bauer (ORCID: https://orcid.org/0000-0002-5740-4386)
- Wenxuan Bao (ORCID: https://orcid.org/0000-0002-1892-9016)
- M Jadhav (ORCID: https://orcid.org/0009-0004-7730-726X)
Institutions
- University of Florida (US)
Publication Details
- Published
- 2025-05-15
- DOI
- https://doi.org/10.1109/spw67851.2025.00020
- Citations
- 1
- Primary Topic
- Speech Recognition and Synthesis
- Type
- article
- Field-Weighted Citation Impact
- 2.48
Funders
- National Science Foundation