Characterizing Audio Adversarial Examples Using Temporal Dependency

TL;DR

This paper investigates the use of temporal dependency in audio data to improve the detection and robustness of speech recognition systems against adversarial attacks, demonstrating its effectiveness over existing methods.

Contribution

It introduces the exploitation of temporal dependency as a novel defense mechanism against audio adversarial examples, showing its resistance to adaptive attacks.

Findings

Temporal dependency enhances discrimination against adversarial examples.

Input transformation methods offer limited robustness improvements.

Temporal dependency-based methods resist advanced adaptive attacks.

Abstract

Recent studies have highlighted adversarial examples as a ubiquitous threat to different neural network models and many downstream applications. Nonetheless, as unique data properties have inspired distinct and powerful learning principles, this paper aims to explore their potentials towards mitigating adversarial inputs. In particular, our results reveal the importance of using the temporal dependency in audio data to gain discriminate power against adversarial examples. Tested on the automatic speech recognition (ASR) tasks and three recent audio adversarial attacks, we find that (i) input transformation developed from image adversarial defense provides limited robustness improvement and is subtle to advanced attacks; (ii) temporal dependency can be exploited to gain discriminative power against audio adversarial examples and is resistant to adaptive attacks considered in our experiments. Our results not only show promising means of improving the robustness of ASR systems, but also offer novel insights in exploiting domain-specific data properties to mitigate negative effects of adversarial examples.