Adversarial Example Detection by Classification for Deep Speech Recognition

Saeid Samizade, Zheng-Hua Tan, Chao Shen, Guan Xiaohong

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

2 Citations (Scopus)

Abstract

Machine Learning systems are vulnerable to adversarial attacks and will highly likely produce incorrect outputs under these attacks. There are white-box and black-box attacks regarding to adversary’s access level to the victim learning algorithm. To defend the learning systems from these attacks, existing methods in the speech domain focus on modifying input signals and testing the behaviours of speech recognizers. We, however, formulate the defense as a classification problem and present a strategy for systematically generating adversarial example datasets: one for white-box attacks and one for black-box attacks, containing both adversarial and normal examples. The white-box attack is a gradient-based method on Baidu DeepSpeech with the Mozilla Common Voice database while the black-box attack is a gradient-free method on a deep model-based keyword spotting system with the Google Speech Command dataset. The generated datasets are used to train a proposed Convolutional Neural Network (CNN), together with cepstral features, to detect adversarial examples. Experimental results show that, it is possible to accurately distinct between adversarial and normal examples for known attacks, in both single-condition and multi-condition training settings, while the performance degrades dramatically for unknown attacks. The adversarial datasets and the source code are made publicly available.
Original languageEnglish
Title of host publicationICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Number of pages5
PublisherIEEE
Publication date9 Apr 2020
Pages3102-3106
Article number9054750
ISBN (Print)978-1-5090-6632-2
ISBN (Electronic)978-1-5090-6631-5
DOIs
Publication statusPublished - 9 Apr 2020
EventICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) - Barcelona, Spain
Duration: 4 May 20208 May 2020

Conference

ConferenceICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
CountrySpain
CityBarcelona
Period04/05/202008/05/2020
SeriesICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
ISSN1520-6149

Keywords

  • Adversarial attack
  • Cepstral feature
  • Convolutional neural network
  • Speech recognition

Fingerprint Dive into the research topics of 'Adversarial Example Detection by Classification for Deep Speech Recognition'. Together they form a unique fingerprint.

Cite this