The vulnerability of deep neural networks to small, adversarially designed perturbations can be attributed to their "excessive linearity." In this paper, we propose a bottom-up strategy for attenuating adversarial perturbations using a nonlinear front end which polarizes and quantizes the data. We observe that ideal polarization can be utilized to completely eliminate perturbations, develop algorithms to learn approximately polarizing bases for data, and investigate the effectiveness of the proposed strategy on the MNIST and Fashion MNIST datasets.
@article{arxiv.2002.09580,
title = {Polarizing Front Ends for Robust CNNs},
author = {Can Bakiskan and Soorya Gopalakrishnan and Metehan Cekic and Upamanyu Madhow and Ramtin Pedarsani},
journal= {arXiv preprint arXiv:2002.09580},
year = {2020}
}
Comments
Published in 45th International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2020)