PATE-AAE: Incorporating Adversarial Autoencoder into Private Aggregation of Teacher Ensembles for Spoken Command Classification

Abstract

We propose using an adversarial autoencoder (AAE) to replace generative adversarial network (GAN) in the private aggregation of teacher ensembles (PATE), a solution for ensuring differential privacy in speech applications. The AAE architecture allows us to obtain good synthetic speech leveraging upon a discriminative training of latent vectors. Such synthetic speech is used to build a privacy-preserving classifier when non-sensitive data is not sufficiently available in the public domain. This classifier follows the PATE scheme that uses an ensemble of noisy outputs to label the synthetic samples and guarantee $\epsilon$-differential privacy (DP) on its derived classifiers. Our proposed framework thus consists of an AAE-based generator and a PATE-based classifier (PATE-AAE). Evaluated on the Google Speech Commands Dataset Version II, the proposed PATE-AAE improves the average classification accuracy by +$2.11\%$ and +$6.60\%$, respectively, when compared with alternative privacy-preserving solutions, namely PATE-GAN and DP-GAN, while maintaining a strong level of privacy target at $\epsilon$=0.01 with a fixed $\delta$=10${}^{-5}$.

arXiv commentAccepted to Interspeech 2021

References in corpus (12)

• Deep Learning with Differential Privacy
• Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition
• Scalable Private Learning with PATE
• Differentially Private Generative Adversarial Network
• GAN-Leaks: A Taxonomy of Membership Inference Attacks against Generative Models
• Extracting Training Data from Large Language Models
• On Convergence and Stability of GANs
• Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data
• Characterizing Speech Adversarial Examples Using Self-Attention U-Net Enhancement
• GS-WGAN: A Gradient-Sanitized Approach for Learning Differentially Private Generators
• Guidelines for Implementing and Auditing Differentially Private Systems
• Submodular Rank Aggregation on Score-based Permutations for Distributed Automatic Speech Recognition

Cited by in corpus (3)

• Voice2Series: Reprogramming Acoustic Models for Time Series Classification
• Longer Version for "Deep Context-Encoding Network for Retinal Image Captioning"
• Robust Unsupervised Multi-Object Tracking in Noisy Environments