Improving Membership Inference in ASR Model Auditing with Perturbed Loss Features

TL;DR

This paper introduces a novel loss-based feature approach with perturbations to improve membership inference attacks on ASR models, enhancing privacy auditing capabilities.

Contribution

It proposes a new combination of loss-based features with Gaussian and adversarial perturbations for MI in ASR, a method not previously explored.

Findings

Loss-based features outperform error-based features in sample-level MI.

Perturbed loss features improve speaker-level MI results.

Different feature sets and access levels significantly impact MI effectiveness.

Abstract

Membership Inference (MI) poses a substantial privacy threat to the training data of Automatic Speech Recognition (ASR) systems, while also offering an opportunity to audit these models with regard to user data. This paper explores the effectiveness of loss-based features in combination with Gaussian and adversarial perturbations to perform MI in ASR models. To the best of our knowledge, this approach has not yet been investigated. We compare our proposed features with commonly used error-based features and find that the proposed features greatly enhance performance for sample-level MI. For speaker-level MI, these features improve results, though by a smaller margin, as error-based features already obtained a high performance for this task. Our findings emphasise the importance of considering different feature sets and levels of access to target models for effective MI in ASR systems, providing valuable insights for auditing such models.