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

01/05/2018

∙

by   Nicholas Carlini, et al.

∙

0

∙

share

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