Understanding the Tradeoffs in Client-side Privacy for Downstream Speech Tasks

Peter Wu*, Paul Pu Liang, Jiatong Shi, Ruslan Salakhutdinov, Shinji Watanabe, Louis Philippe Morency

*この研究の対応する著者

研究成果: Conference contribution

抄録

As users increasingly rely on cloud-based computing services, it is important to ensure that uploaded speech data re-mains private. Existing solutions rely either on server-side meth-ods or focus on hiding speaker identity. While these approaches reduce certain security concerns, they do not give users client-side control over whether their biometric information is sent to the server. In this paper, we formally define client-side privacy and discuss its unique technical challenges requiring 1) direct manipulation of raw data on client devices, 2) adaptability with a broad range of server-side processing models, and 3) low time and space complexity for compatibility with limited-bandwidth devices. These unique challenges require a new class of models that achieve fidelity in reconstruction, privacy preservation of sensitive personal attributes, and efficiency during training and inference. As a step towards client-side privacy for speech recog-nition, we investigate three techniques spanning signal processing, disentangled representation learning, and adversarial training. Through a series gender and accent masking tasks, we observe that each method has its unique strengths, but none manage to effectively balance the trade-offs between performance, privacy, and complexity. These insights call for more research in client-side privacy to ensure a safer deployment of cloud-based speech processing services.

本文言語English
ホスト出版物のタイトル2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2021 - Proceedings
出版社Institute of Electrical and Electronics Engineers Inc.
ページ841-848
ページ数8
ISBN(電子版)9789881476890
出版ステータスPublished - 2021
外部発表はい
イベント2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2021 - Tokyo, Japan
継続期間: 2021 12月 142021 12月 17

出版物シリーズ

名前2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2021 - Proceedings

Conference

Conference2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2021
国/地域Japan
CityTokyo
Period21/12/1421/12/17

ASJC Scopus subject areas

  • 人工知能
  • コンピュータ ビジョンおよびパターン認識
  • 信号処理
  • 器械工学

フィンガープリント

「Understanding the Tradeoffs in Client-side Privacy for Downstream Speech Tasks」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル