Cortical Features for Defense Against Adversarial Audio Attacks
arXiv:2102.00313
Abstract
We propose using a computational model of the auditory cortex as a defense against adversarial attacks on audio. We apply several white-box iterative optimization-based adversarial attacks to an implementation of Amazon Alexa's HW network, and a modified version of this network with an integrated cortical representation, and show that the cortical features help defend against universal adversarial examples. At the same level of distortion, the adversarial noises found for the cortical network are always less effective for universal audio attacks. We make our code publicly available at https://github.com/ilyakava/py3fst.
arXiv commentCo-author legal name changed
References in corpus (7)
- Pyroomacoustics: A Python package for audio room simulations and array processing algorithms
- Imperceptible, Robust, and Targeted Adversarial Examples for Automatic Speech Recognition
- Joint Time-Frequency Scattering for Audio Classification
- Universal Adversarial Audio Perturbations
- Adversarial Music: Real World Audio Adversary Against Wake-word Detection System
- Perceptual Based Adversarial Audio Attacks
- Small-Footprint Open-Vocabulary Keyword Spotting with Quantized LSTM Networks