Guardado en:
Detalles Bibliográficos
Autores principales: Zhao, Kaikang, Chen, Xi, Huang, Wei, Ding, Liuxin, Kong, Xianglong, Zhang, Fan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2403.16405
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866929287561478144
author Zhao, Kaikang
Chen, Xi
Huang, Wei
Ding, Liuxin
Kong, Xianglong
Zhang, Fan
author_facet Zhao, Kaikang
Chen, Xi
Huang, Wei
Ding, Liuxin
Kong, Xianglong
Zhang, Fan
contents The integration of an ensemble of deep learning models has been extensively explored to enhance defense against adversarial attacks. The diversity among sub-models increases the attack cost required to deceive the majority of the ensemble, thereby improving the adversarial robustness. While existing approaches mainly center on increasing diversity in feature representations or dispersion of first-order gradients with respect to input, the limited correlation between these diversity metrics and adversarial robustness constrains the performance of ensemble adversarial defense. In this work, we aim to enhance ensemble diversity by reducing attack transferability. We identify second-order gradients, which depict the loss curvature, as a key factor in adversarial robustness. Computing the Hessian matrix involved in second-order gradients is computationally expensive. To address this, we approximate the Hessian-vector product using differential approximation. Given that low curvature provides better robustness, our ensemble model was designed to consider the influence of curvature among different sub-models. We introduce a novel regularizer to train multiple more-diverse low-curvature network models. Extensive experiments across various datasets demonstrate that our ensemble model exhibits superior robustness against a range of attacks, underscoring the effectiveness of our approach.
format Preprint
id arxiv_https___arxiv_org_abs_2403_16405
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Ensemble Adversarial Defense via Integration of Multiple Dispersed Low Curvature Models
Zhao, Kaikang
Chen, Xi
Huang, Wei
Ding, Liuxin
Kong, Xianglong
Zhang, Fan
Machine Learning
Cryptography and Security
Computer Vision and Pattern Recognition
The integration of an ensemble of deep learning models has been extensively explored to enhance defense against adversarial attacks. The diversity among sub-models increases the attack cost required to deceive the majority of the ensemble, thereby improving the adversarial robustness. While existing approaches mainly center on increasing diversity in feature representations or dispersion of first-order gradients with respect to input, the limited correlation between these diversity metrics and adversarial robustness constrains the performance of ensemble adversarial defense. In this work, we aim to enhance ensemble diversity by reducing attack transferability. We identify second-order gradients, which depict the loss curvature, as a key factor in adversarial robustness. Computing the Hessian matrix involved in second-order gradients is computationally expensive. To address this, we approximate the Hessian-vector product using differential approximation. Given that low curvature provides better robustness, our ensemble model was designed to consider the influence of curvature among different sub-models. We introduce a novel regularizer to train multiple more-diverse low-curvature network models. Extensive experiments across various datasets demonstrate that our ensemble model exhibits superior robustness against a range of attacks, underscoring the effectiveness of our approach.
title Ensemble Adversarial Defense via Integration of Multiple Dispersed Low Curvature Models
topic Machine Learning
Cryptography and Security
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.16405