arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2001.08444eess.AScs.HCcs.LGcs.SDstat.ML
收藏

音频对抗示例的人工评估

On the human evaluation of audio adversarial examples

Faculty of Informatics(信息学院);University of the Basque Country (UPV/EHU)(巴斯克大学)

Jon Vadillo, Roberto Santana

摘要:

人机交互越来越依赖于语音通信。机器学习模型通常用于解释人类语音命令。然而,这些模型可能会被对抗性的例子所愚弄,这些例子是故意扰动的输入,以产生错误的预测而不被注意到。虽然许多研究都集中在开发新技术来产生对抗性扰动,但较少关注确定人类是否以及如何注意到扰动的方面。这个问题是相关的,因为提出的对抗扰动策略的高欺骗率只有在扰动不可检测的情况下才有价值。在本文中,我们调查在何种程度上的失真度量在文献中提出的音频对抗性的例子,这是通常应用于评估的有效性的方法来生成这些攻击,是一个可靠的措施,人类感知的扰动。使用一个分析框架,和一个实验中,18名受试者评估音频对抗的例子,我们证明,公约所采用的指标是不是一个可靠的措施的感知相似性的对抗的例子在音频域。

英文摘要:

Human-machine interaction is increasingly dependent on speech communication. Machine Learning models are usually applied to interpret human speech commands. However, these models can be fooled by adversarial examples, which are inputs intentionally perturbed to produce a wrong prediction without being noticed. While much research has been focused on developing new techniques to generate adversarial perturbations, less attention has been given to aspects that determine whether and how the perturbations are noticed by humans. This question is relevant since high fooling rates of proposed adversarial perturbation strategies are only valuable if the perturbations are not detectable. In this paper we investigate to which extent the distortion metrics proposed in the literature for audio adversarial examples, and which are commonly applied to evaluate the effectiveness of methods for generating these attacks, are a reliable measure of the human perception of the perturbations. Using an analytical framework, and an experiment in which 18 subjects evaluate audio adversarial examples, we demonstrate that the metrics employed by convention are not a reliable measure of the perceptual similarity of adversarial examples in the audio domain.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑