Inference Attacks for X-Vector Speaker Anonymization

TL;DR

This paper introduces a simple, machine learning-free inference attack on x-vector speaker anonymization that outperforms existing complex attack methods, highlighting privacy vulnerabilities.

Contribution

The paper presents a novel, ML-free inference attack method that surpasses existing approaches in de-anonymizing x-vector speaker anonymization systems.

Findings

The attack outperforms existing methods in de-anonymization accuracy.

It demonstrates vulnerabilities in current speaker anonymization techniques.

The approach is simpler and more effective than previous ML-based attacks.

Abstract

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.