Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2509.14469

Measuring Soft Biometric Leakage in Speaker De-Identification Systems

Seungmin Seo·Oleg Aulov·P. Jonathon Phillips
View original source ↗Read source document ↗Export metadata ↓

Abstract

We use the term re-identification to refer to the process of recovering the original speaker's identity from anonymized speech outputs. Speaker de-identification systems aim to reduce the risk of re-identification, but most evaluations focus only on individual-level measures and overlook broader risks from soft biometric leakage. We introduce the Soft Biometric Leakage Score (SBLS), a unified method that quantifies resistance to zero-shot inference attacks on non-unique traits such as channel type, age range, dialect, sex of the speaker, or speaking style. SBLS integrates three elements: direct attribute inference using pre-trained classifiers, linkage detection via mutual information analysis, and subgroup robustness across intersecting attributes. Applying SBLS with publicly available classifiers, we show that all five evaluated de-identification systems exhibit significant vulnerabilities. Our results indicate that adversaries using only pre-trained models - without access to original speech or system details - can still reliably recover soft biometric information from anonymized output, exposing fundamental weaknesses that standard distributional metrics fail to capture.

Explore related subjects

cs.SD

Keep this discovery

BibTeXRIS

Seungmin Seo, Oleg Aulov, P. Jonathon Phillips. 2025-09-17. Measuring Soft Biometric Leakage in Speaker De-Identification Systems. https://arxiv.org/abs/2509.14469

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Xiaomi-CocktailASR-1 Technical Report

Recently, large language model (LLM) based ASR models have achieved significant progress, yet they generally lack support for multi-speaker scenarios, where the cocktail party problem remains a critical bottleneck for further advancing ASR. Existing TS-ASR methods, including end-to-end architectures with speaker embeddings and latest LLM-based explorations suffer from degraded single-speaker performance and the inability to reject when the target speaker is absent. In this paper, we propose Xiaomi-CocktailASR-1, an LLM-based end-to-end TS-ASR architecture. By utilizing reference speech as voiceprint prompts, it directly transcribes the target speaker's speech without requiring speech separation. Xiaomi-CocktailASR-1 maintains competitive performance in single-speaker scenarios, comparable to mainstream ASR models. It also features a negative sample rejection capability, outputting empty text when the target speaker is absent from the mixed speech. Additionally, Xiaomi-CocktailASR-1 supports a Chain-of-Thought (CoT) reasoning mode to provide explicit reasoning steps. Extensive experiments on various synthetic and real-world multispeaker benchmarks demonstrate that Xiaomi-CocktailASR-1 achieves state-of-the-art performance, effectively addressing the cocktail party problem through a unified architecture that balances multispeaker and single-speaker recognition accuracy, along with rejection capability.

cs.SD↗
arXiv2026

EConv-TasNet: Efficient Conv-TasNet for Effective Speech Separation

Conv-TasNet has served as a strong baseline for time-domain speech separation, and many studies have extended it with advanced architectures such as dual-path networks, U-Nets, and attention mechanisms. However, these methods often introduce high computational cost and complexity, limiting their deployment in resource-constrained scenarios. To address this issue, we propose eConv-TasNet, an efficient variant of Conv-TasNet that improves both effectiveness and efficiency without relying on resource-intensive modules. The proposed model consists of a group-wise early-splitting (GES) module and a multi-group feature aggregation (MGFA) module. GES generates discriminative speaker embeddings at intermediate stages, while MGFA progressively aggregates these group-level representations for refined mask estimation. Experimental results show that eConv-TasNet reduces model size by 22.4%, accelerates inference by 18.9%, and improves SI-SNRi by 14.0%-28.0% across three public benchmarks. Moreover, it achieves competitive performance compared with state-of-the-art methods while requiring significantly fewer parameters and lower inference cost. These results demonstrate a favorable efficiency-effectiveness trade-off for edge deployment.

cs.SD↗