Generating gender-ambiguous voices for privacy-preserving speech recognition

TL;DR

This paper introduces GenGAN, a generative adversarial network that synthesizes gender-ambiguous voices to enhance privacy in speech recognition by concealing private attributes like gender.

Contribution

The paper presents a novel GAN-based method that improves privacy-utility trade-offs in speech recognition by generating gender-ambiguous voices conditioned on gender information.

Findings

GenGAN effectively conceals gender information in synthesized voices.

It outperforms existing privacy-preserving methods in balancing privacy and utility.

The approach maintains speech recognition accuracy while protecting private attributes.

Abstract

Our voice encodes a uniquely identifiable pattern which can be used to infer private attributes, such as gender or identity, that an individual might wish not to reveal when using a speech recognition service. To prevent attribute inference attacks alongside speech recognition tasks, we present a generative adversarial network, GenGAN, that synthesises voices that conceal the gender or identity of a speaker. The proposed network includes a generator with a U-Net architecture that learns to fool a discriminator. We condition the generator only on gender information and use an adversarial loss between signal distortion and privacy preservation. We show that GenGAN improves the trade-off between privacy and utility compared to privacy-preserving representation learning methods that consider gender information as a sensitive attribute to protect.