[1911.00126] Adversarial Music: Real World Audio Adversary Against Wake-word Detection System

Voice Assistants (VAs) such as Amazon Alexa or Google Assistant rely on wake-word detection to respond to people's commands, which could potentially be vulnerable to audio adversarial examples. In this work, we target our attack on the wake-word detection system, jamming the model with some inconspicuous background music to deactivate the VAs while our audio adversary is present. We implemented an emulated wake-word detection system of Amazon Alexa based on recent publications. We validated our models against the real Alexa in terms of wake-word detection accuracy. Then we computed our audio adversaries with consideration of expectation over transform and we implemented our audio adversary with a differentiable synthesizer. Next, we verified our audio adversaries digitally on hundreds of samples of utterances collected from the real world. Our experiments show that we can effectively reduce the recognition F1 score of our emulated model from 93.4% to 11.0%. Finally, we tested our audio

4 mentions: @jonathanfly@bbr_bbq@cynicalsecurity@iansample
Date: 2019/11/07 23:20

Referring Tweets

@bbr_bbq 細工したBGMを流すことで、Wake-Wordによる音声アシスタントの操作をジャミングする手法。検証では、Alexaの音声認識精度を92.5%から11.0%にまで下げる事に成功したとのこと。 Adversarial Music: Real World Audio Adversary Against Wake-word Detection System t.co/SyY4kq3rSx
@jonathanfly Adversarial Music that makes it hard for Alexa to wake up. They didn't test with music played on Alexa itself, wonder if you could publish an hour-long version on Amazon Music and ask a friend to request it...🤔 abstract: t.co/DNfM3skZ9R videos: t.co/jqsz5rbCXR t.co/99whsgVCJr

Related Entries

Read more Adversarial Examples 分野の動向 (敵対的サンプル発表資料)
1 users, 4 mentions 2019/09/19 03:48
Read more Adversarial Examples Are Not Bugs, They Are Featuresを読んだのでメモ - 機械学習とかコンピュータビジョンとか
1 users, 0 mentions 2019/09/21 07:48
Read more Deep Learning Adversarial Examples – Clarifying Misconceptions
0 users, 0 mentions 2018/04/22 03:40
Read more EXPLAINING AND HARNESSING ADVERSARIAL EXAMPLESを読んだのでメモ - 機械学習とかコンピュータビジョンとか
1 users, 0 mentions 2019/09/17 10:21
Read more GitHub - pfnet-research/nips17-adversarial-attack: Submission to Kaggle NIPS'17 competition on adver...
0 users, 0 mentions 2018/04/22 03:40