Guardado en:
Detalles Bibliográficos
Autores principales: Singh, Himanshu, Subramanyam, A. V., Rajput, Shivank, Kankanhalli, Mohan
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2509.07673
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913017341411328
author Singh, Himanshu
Subramanyam, A. V.
Rajput, Shivank
Kankanhalli, Mohan
author_facet Singh, Himanshu
Subramanyam, A. V.
Rajput, Shivank
Kankanhalli, Mohan
contents Deep neural networks have exhibited impressive performance in image classification tasks but remain vulnerable to adversarial examples. Standard adversarial training enhances robustness but typically fails to explicitly address inter-class feature overlap, a significant contributor to adversarial susceptibility. In this work, we introduce a novel adversarial training framework that actively mitigates inter-class proximity by projecting out inter-class dependencies from adversarial and clean samples in the feature space. Specifically, our approach first identifies the nearest inter-class neighbors for each adversarial sample and subsequently removes projections onto these neighbors to enforce stronger feature separability. Theoretically, we demonstrate that our proposed logits correction reduces the Lipschitz constant of neural networks, thereby lowering the Rademacher complexity, which directly contributes to improved generalization and robustness. Extensive experiments across standard benchmarks including CIFAR-10, CIFAR-100, SVHN, and TinyImagenet show that our method demonstrates strong performance that is competitive with leading adversarial training techniques, highlighting significant achievements in both robust and clean accuracy. Our findings reveal the importance of addressing inter-class feature proximity explicitly to bolster adversarial robustness in DNNs. The code is available in the supplementary material.
format Preprint
id arxiv_https___arxiv_org_abs_2509_07673
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Nearest Neighbor Projection Removal Adversarial Training
Singh, Himanshu
Subramanyam, A. V.
Rajput, Shivank
Kankanhalli, Mohan
Computer Vision and Pattern Recognition
Machine Learning
68T45 (Primary), 68T10 (Secondary)
I.5.4
Deep neural networks have exhibited impressive performance in image classification tasks but remain vulnerable to adversarial examples. Standard adversarial training enhances robustness but typically fails to explicitly address inter-class feature overlap, a significant contributor to adversarial susceptibility. In this work, we introduce a novel adversarial training framework that actively mitigates inter-class proximity by projecting out inter-class dependencies from adversarial and clean samples in the feature space. Specifically, our approach first identifies the nearest inter-class neighbors for each adversarial sample and subsequently removes projections onto these neighbors to enforce stronger feature separability. Theoretically, we demonstrate that our proposed logits correction reduces the Lipschitz constant of neural networks, thereby lowering the Rademacher complexity, which directly contributes to improved generalization and robustness. Extensive experiments across standard benchmarks including CIFAR-10, CIFAR-100, SVHN, and TinyImagenet show that our method demonstrates strong performance that is competitive with leading adversarial training techniques, highlighting significant achievements in both robust and clean accuracy. Our findings reveal the importance of addressing inter-class feature proximity explicitly to bolster adversarial robustness in DNNs. The code is available in the supplementary material.
title Nearest Neighbor Projection Removal Adversarial Training
topic Computer Vision and Pattern Recognition
Machine Learning
68T45 (Primary), 68T10 (Secondary)
I.5.4
url https://arxiv.org/abs/2509.07673