# Improving the Speaker Anonymization Evaluation's Robustness to Target Speakers with Adversarial Learning

arXiv: https://arxiv.org/abs/2508.09803 · HTML: https://mlsanity.com/paper/2508.09803/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 18 references that are in the map.
- First posted on arXiv 13 Aug 2025.

## Builds on (most cited first, 12 of 12)

- [Unsupervised Domain Adaptation by Backpropagation](https://mlsanity.com/paper/1409.7495/) (2014)
- [LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech](https://mlsanity.com/paper/1904.02882/) (2019)
- [An Overview of Voice Conversion and its Challenges: From Statistical Modeling to Deep Learning](https://mlsanity.com/paper/2008.03648/) (2020)
- [Voice Conversion With Just Nearest Neighbors](https://mlsanity.com/paper/2305.18975/) (2023)
- [Introducing the VoicePrivacy Initiative](https://mlsanity.com/paper/2005.01387/) (2020)
- [ECAPA-TDNN Embeddings for Speaker Diarization](https://mlsanity.com/paper/2104.01466/) (2021)
- [The VoicePrivacy 2024 Challenge Evaluation Plan](https://mlsanity.com/paper/2404.02677/) (2024)
- [Domain Adversarial Training for Accented Speech Recognition](https://mlsanity.com/paper/1806.02786/) (2018)
- [Anonymizing Speech: Evaluating and Designing Speaker Anonymization Techniques](https://mlsanity.com/paper/2308.04455/) (2023)
- [Noise Adaptive Speech Enhancement using Domain Adversarial Training](https://mlsanity.com/paper/1807.07501/) (2018)
- [Private kNN-VC: Interpretable Anonymization of Converted Speech](https://mlsanity.com/paper/2505.17584/) (2025)
- [The Risks and Detection of Overestimated Privacy Protection in Voice Anonymisation](https://mlsanity.com/paper/2507.22534/) (2025)

## Abstract (arXiv)

The current privacy evaluation for speaker anonymization often overestimates privacy when a same-gender target selection algorithm (TSA) is used, although this TSA leaks the speaker's gender and should hence be more vulnerable. We hypothesize that this occurs because the evaluation does not account for the fact that anonymized speech contains information from both the source and target speakers. To address this, we propose to add a target classifier that measures the influence of target speaker information in the evaluation, which can also be removed with adversarial learning. Experiments demonstrate that this approach is effective for multiple anonymizers, particularly when using a same-gender TSA, leading to a more reliable assessment.