Boosting the Transferability of Audio Adversarial Examples with Acoustic Representation Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jin, Weifei, Su, Junjie, Wang, Hejia, Ye, Yulin, Hao, Jie
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916662147547136
author Jin, Weifei
Su, Junjie
Wang, Hejia
Ye, Yulin
Hao, Jie
author_facet Jin, Weifei
Su, Junjie
Wang, Hejia
Ye, Yulin
Hao, Jie
contents With the widespread application of automatic speech recognition (ASR) systems, their vulnerability to adversarial attacks has been extensively studied. However, most existing adversarial examples are generated on specific individual models, resulting in a lack of transferability. In real-world scenarios, attackers often cannot access detailed information about the target model, making query-based attacks unfeasible. To address this challenge, we propose a technique called Acoustic Representation Optimization that aligns adversarial perturbations with low-level acoustic characteristics derived from speech representation models. Rather than relying on model-specific, higher-layer abstractions, our approach leverages fundamental acoustic representations that remain consistent across diverse ASR architectures. By enforcing an acoustic representation loss to guide perturbations toward these robust, lower-level representations, we enhance the cross-model transferability of adversarial examples without degrading audio quality. Our method is plug-and-play and can be integrated with any existing attack methods. We evaluate our approach on three modern ASR models, and the experimental results demonstrate that our method significantly improves the transferability of adversarial examples generated by previous methods while preserving the audio quality.
format Preprint
id arxiv_https___arxiv_org_abs_2503_19591
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Boosting the Transferability of Audio Adversarial Examples with Acoustic Representation Optimization
Jin, Weifei
Su, Junjie
Wang, Hejia
Ye, Yulin
Hao, Jie
Sound
Cryptography and Security
Machine Learning
Audio and Speech Processing
With the widespread application of automatic speech recognition (ASR) systems, their vulnerability to adversarial attacks has been extensively studied. However, most existing adversarial examples are generated on specific individual models, resulting in a lack of transferability. In real-world scenarios, attackers often cannot access detailed information about the target model, making query-based attacks unfeasible. To address this challenge, we propose a technique called Acoustic Representation Optimization that aligns adversarial perturbations with low-level acoustic characteristics derived from speech representation models. Rather than relying on model-specific, higher-layer abstractions, our approach leverages fundamental acoustic representations that remain consistent across diverse ASR architectures. By enforcing an acoustic representation loss to guide perturbations toward these robust, lower-level representations, we enhance the cross-model transferability of adversarial examples without degrading audio quality. Our method is plug-and-play and can be integrated with any existing attack methods. We evaluate our approach on three modern ASR models, and the experimental results demonstrate that our method significantly improves the transferability of adversarial examples generated by previous methods while preserving the audio quality.
title Boosting the Transferability of Audio Adversarial Examples with Acoustic Representation Optimization
topic Sound
Cryptography and Security
Machine Learning
Audio and Speech Processing
url https://arxiv.org/abs/2503.19591