Masked Spatial-Spectral Autoencoders Are Excellent Hyperspectral Defenders

Qi, Jiahao; Gong, Zhiqiang; Liu, Xingyue; Bin, Kangcheng; Chen, Chen; Li, Yongqian; Xue, Wei; Zhong, Ping

doi:10.48550/arxiv.2207.07803

Cited by 1 publication

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors in [11] use their Masked Spatial-Spectral Autoencoder (MSSA) which consists of masked sequence attention learning, dynamic graph embedding, and self-supervised re- construction. Rather than focusing on semantic segmentation via network architecture, some works like [12,13] try to use the rich spectral information to robustify the entire process. In [12], they propose a spectral sampling and shape encoding to increase adversarial robustness as a preprocessing step to traditional per-pixel classification via random sampling.…”

Section: Related Workmentioning

confidence: 99%

“…In [12], they propose a spectral sampling and shape encoding to increase adversarial robustness as a preprocessing step to traditional per-pixel classification via random sampling. In work [13], rather than random sampling, they use autoencoders to reconstruct the spectral signature of pixels for later classification with a shared global loss function. However, all these approaches focus on a single attack at a time and do not explore network robustness in the presence of multiple attacks.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Improving Hyperspectral Adversarial Robustness using Ensemble Networks in the Presences of Multiple Attacks

Soucy¹,

Sekeh²

2022

Preprint

View full text Add to dashboard Cite

Semantic segmentation of hyperspectral images (HSI) has seen great strides in recent years by incorporating knowledge from deep learning RGB classification models. Similar to their classification counterparts, semantic segmentation models are vulnerable to adversarial examples and need adversarial training to counteract them. 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 the 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. Our approach allows for the presence of multiple attacks mixed together while also labeling attack types during testing. We experimentally show that ADE-Net outperforms the baseline, which is a single network adversarially trained under a mix of multiple attacks, for HSI Indian Pines, Kennedy Space, and Houston datasets.

show abstract