Exploring Audio Editing Features as User-Centric Privacy Defenses Against Large Language Model(LLM) Based Emotion Inference Attacks

TL;DR

This paper proposes a user-centric audio editing approach using pitch and tempo manipulation to protect emotional privacy in speech data, effectively countering LLM-based inference attacks while maintaining usability.

Contribution

It introduces a practical privacy defense leveraging familiar audio editing features, validated against diverse adversarial models and designed for on-device implementation.

Findings

Pitch and tempo manipulation effectively obfuscate emotional data.

The approach maintains usability and is applicable across Android and iOS.

Effective against DNN, LLM, and reversibility attacks.

Abstract

The rapid proliferation of speech-enabled technologies, including virtual assistants, video conferencing platforms, and wearable devices, has raised significant privacy concerns, particularly regarding the inference of sensitive emotional information from audio data. Existing privacy-preserving methods often compromise usability and security, limiting their adoption in practical scenarios. This paper introduces a novel, user-centric approach that leverages familiar audio editing techniques, specifically pitch and tempo manipulation, to protect emotional privacy without sacrificing usability. By analyzing popular audio editing applications on Android and iOS platforms, we identified these features as both widely available and usable. We rigorously evaluated their effectiveness against a threat model, considering adversarial attacks from diverse sources, including Deep Neural Networks (DNNs), Large Language Models (LLMs), and and reversibility testing. Our experiments, conducted on three distinct datasets, demonstrate that pitch and tempo manipulation effectively obfuscates emotional data. Additionally, we explore the design principles for lightweight, on-device implementation to ensure broad applicability across various devices and platforms.