Whisper Smarter, not Harder: Adversarial Attack on Partial Suppression

TL;DR

This paper examines adversarial attacks on speech recognition models, demonstrating that partial suppression attacks are more imperceptible and exploring defenses like low-pass filtering to improve robustness.

Contribution

It introduces the concept of partial suppression in adversarial attacks on ASR models and evaluates its effectiveness and defenses.

Findings

Partial suppression attacks are more imperceptible than complete suppression.

Low-pass filter can serve as an effective defense against these attacks.

Relaxing optimization objectives enhances attack stealthiness.

Abstract

Currently, Automatic Speech Recognition (ASR) models are deployed in an extensive range of applications. However, recent studies have demonstrated the possibility of adversarial attack on these models which could potentially suppress or disrupt model output. We investigate and verify the robustness of these attacks and explore if it is possible to increase their imperceptibility. We additionally find that by relaxing the optimisation objective from complete suppression to partial suppression, we can further decrease the imperceptibility of the attack. We also explore possible defences against these attacks and show a low-pass filter defence could potentially serve as an effective defence.