Audio Adversarial Examples: Targeted Attacks on Speech-to-Text

被引：533

作者：

Carlini, Nicholas ^{[1
]}

Wagner, David ^{[1
]}

机构：

[1] Univ Calif Berkeley, Berkeley, CA 94720 USA

来源：

2018 IEEE SYMPOSIUM ON SECURITY AND PRIVACY WORKSHOPS (SPW 2018) | 2018年

基金：

美国国家科学基金会;

关键词：

D O I：

10.1109/SPW.2018.00009

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

We construct targeted audio adversarial examples on automatic speech recognition. Given any audio waveform, we can produce another that is over 99.9% similar, but transcribes as any phrase we choose (recognizing up to 50 characters per second of audio). We apply our white-box iterative optimization-based attack to Mozilla's implementation DeepSpeech end-to-end, and show it has a 100% success rate. The feasibility of this attack introduce a new domain to study adversarial examples.

引用

页码：1 / 7

页数：7