Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Examining the Interplay Between Privacy and Fairness for Speech Processing: A Review and Perspective

Anna Leschanowsky, Sneha Das

TL;DR

The paper addresses the underexplored interplay between privacy and fairness in speech processing. It develops a lifecycle-based framework that links privacy harms and biases, using Solove's taxonomy and Suresh's ML lifecycle to map risks across data collection, model building, and deployment. It reviews how privacy-enhancing technologies (PETs) and bias mitigation can influence each other and documents the paucity of speech-specific studies on privacy–fairness tradeoffs, offering open questions and guidance for evaluation. The work aims to spur integrated privacy-aware and fairness-aware algorithm design in speech technologies with practical implications for sensitive domains such as healthcare, law enforcement, and consumer applications.

Abstract

Speech technology has been increasingly deployed in various areas of daily life including sensitive domains such as healthcare and law enforcement. For these technologies to be effective, they must work reliably for all users while preserving individual privacy. Although tradeoffs between privacy and utility, as well as fairness and utility, have been extensively researched, the specific interplay between privacy and fairness in speech processing remains underexplored. This review and position paper offers an overview of emerging privacy-fairness tradeoffs throughout the entire machine learning lifecycle for speech processing. By drawing on well-established frameworks on fairness and privacy, we examine existing biases and sources of privacy harm that coexist during the development of speech processing models. We then highlight how corresponding privacy-enhancing technologies have the potential to inadvertently increase these biases and how bias mitigation strategies may conversely reduce privacy. By raising open questions, we advocate for a comprehensive evaluation of privacy-fairness tradeoffs for speech technology and the development of privacy-enhancing and fairness-aware algorithms in this domain.

Examining the Interplay Between Privacy and Fairness for Speech Processing: A Review and Perspective

TL;DR

The paper addresses the underexplored interplay between privacy and fairness in speech processing. It develops a lifecycle-based framework that links privacy harms and biases, using Solove's taxonomy and Suresh's ML lifecycle to map risks across data collection, model building, and deployment. It reviews how privacy-enhancing technologies (PETs) and bias mitigation can influence each other and documents the paucity of speech-specific studies on privacy–fairness tradeoffs, offering open questions and guidance for evaluation. The work aims to spur integrated privacy-aware and fairness-aware algorithm design in speech technologies with practical implications for sensitive domains such as healthcare, law enforcement, and consumer applications.

Abstract

Speech technology has been increasingly deployed in various areas of daily life including sensitive domains such as healthcare and law enforcement. For these technologies to be effective, they must work reliably for all users while preserving individual privacy. Although tradeoffs between privacy and utility, as well as fairness and utility, have been extensively researched, the specific interplay between privacy and fairness in speech processing remains underexplored. This review and position paper offers an overview of emerging privacy-fairness tradeoffs throughout the entire machine learning lifecycle for speech processing. By drawing on well-established frameworks on fairness and privacy, we examine existing biases and sources of privacy harm that coexist during the development of speech processing models. We then highlight how corresponding privacy-enhancing technologies have the potential to inadvertently increase these biases and how bias mitigation strategies may conversely reduce privacy. By raising open questions, we advocate for a comprehensive evaluation of privacy-fairness tradeoffs for speech technology and the development of privacy-enhancing and fairness-aware algorithms in this domain.
Paper Structure (19 sections, 1 figure)

This paper contains 19 sections, 1 figure.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Bias and Fairness
  4. Privacy and Privacy Harms
  5. Privacy-Fairness Trade-Off in Other Domains
  6. Sources of Privacy Harms and Bias in Speech Processing Tasks
  7. Data Collection and Preparation
  8. Sources of bias during data curation
  9. Sources of privacy harm during data curation
  10. Model Building, Evaluation and Post-Processing
  11. Sources of bias during model building and evaluation
  12. Privacy harm during model building and evaluation
  13. Model Deployment
  14. Sources of privacy harm during deployment
  15. Contextualizing Privacy-Fairness Tradeoffs for Speech Processing
  16. ...and 4 more sections

Figures (1)

  • Figure 1: We depict the general ML lifecycle including associated biases and sources of privacy harms (adapted from suresh2021 and solove2005).