Cortical Features for Defense Against Adversarial Audio Attacks

TL;DR

This paper introduces a cortical feature-based model inspired by the auditory cortex to defend against adversarial audio attacks, demonstrating improved robustness over standard models.

Contribution

The study presents a novel cortical-inspired feature integration into audio neural networks to enhance resistance to universal adversarial examples.

Findings

Cortical features reduce attack effectiveness at the same distortion level.

The cortical model shows improved robustness against universal adversarial audio attacks.

Code implementation is publicly available for replication.

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.