Improving Hyperspectral Adversarial Robustness Under Multiple Attacks

TL;DR

This paper introduces ADE-Net, a unified model for hyperspectral image segmentation that detects attack types and maintains robustness against multiple adversarial attacks simultaneously.

Contribution

The paper proposes ADE-Net, an ensemble network with an attack discriminator to improve robustness against multiple adversarial attacks in hyperspectral image segmentation.

Findings

Enhanced robustness to multiple attacks

Effective attack type detection

Maintains performance across attack types

Abstract

Semantic segmentation models classifying hyperspectral images (HSI) are vulnerable to adversarial examples. Traditional approaches to adversarial robustness focus on training or retraining a single network on attacked data, however, in the presence of multiple attacks these approaches decrease in performance compared to networks trained individually on each attack. To combat this issue we propose an Adversarial Discriminator Ensemble Network (ADE-Net) which focuses on attack type detection and adversarial robustness under a unified model to preserve per data-type weight optimally while robustifiying the overall network. In the proposed method, a discriminator network is used to separate data by attack type into their specific attack-expert ensemble network.