Sparse Coding Frontend for Robust Neural Networks

TL;DR

This paper proposes a novel sparse coding frontend that enhances neural network robustness against various adversarial attacks by attenuating perturbations before classification, trained solely on clean images.

Contribution

Introduction of a sparse coding based frontend as a new defense mechanism against adversarial attacks, differing from traditional adversarial training methods.

Findings

Effective against multiple attack types (Linf, L2, L1)

Significantly reduces adversarial perturbations

Demonstrates robustness on CIFAR-10 dataset

Abstract

Deep Neural Networks are known to be vulnerable to small, adversarially crafted, perturbations. The current most effective defense methods against these adversarial attacks are variants of adversarial training. In this paper, we introduce a radically different defense trained only on clean images: a sparse coding based frontend which significantly attenuates adversarial attacks before they reach the classifier. We evaluate our defense on CIFAR-10 dataset under a wide range of attack types (including Linf , L2, and L1 bounded attacks), demonstrating its promise as a general-purpose approach for defense.