On the Structural Sensitivity of Deep Convolutional Networks to the Directions of Fourier Basis Functions

2018/09/11 by Yusuke Tsuzuku, Tsuzuku, Yusuke, Issei Sato +1 · 2 citations
Computer Science · Engineering · #Advancements in Semiconductor Devices and Circuit Design #Adversarial Robustness in Machine Learning #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Physical Unclonable Functions (PUFs) and Hardware Security

paper · pdf · doi:10.48550/arxiv.1809.04098

openalex publication_date 2018/09/11 · openalex created_date 2019/07/12 · openalex updated_date 2026/07/28

Abstract

Data-agnostic quasi-imperceptible perturbations on inputs are known to degrade recognition accuracy of deep convolutional networks severely. This phenomenon is considered to be a potential security issue. Moreover, some results on statistical generalization guarantees indicate that the phenomenon can be a key to improve the networks' generalization. However, the characteristics of the shared directions of such harmful perturbations remain unknown. Our primal finding is that convolutional networks are sensitive to the directions of Fourier basis functions. We derived the property by specializing a hypothesis of the cause of the sensitivity, known as the linearity of neural networks, to convolutional networks and empirically validated it. As a by-product of the analysis, we propose an algorithm to create shift-invariant universal adversarial perturbations available in black-box settings.

Citations

• An algorithm for the machine calculation of complex Fourier series
• Gradient-based learning applied to document recognition
• ImageNet Large Scale Visual Recognition Challenge
• ImageNet classification with deep convolutional neural networks
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• A study of the effect of JPG compression on adversarial images
• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
• Fast Feature Fool: A data independent approach to universal adversarial\n perturbations
• Art of singular vectors and universal adversarial perturbations
• Robustness of classifiers to universal perturbations: a geometric\n perspective
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Learning Multiple Layers of Features from Tiny Images

Cited by

• Understanding Sensitivity of Differential Attention through the Lens of Adversarial Robustness
• On the Complexity-Faithfulness Trade-off of Gradient-Based Explanations

Related

• Local Convolutions Cause an Implicit Bias towards High Frequency Adversarial Examples
• Geometric robustness of deep networks: analysis and improvement
• Sensitivity of Deep Convolutional Networks to Gabor Noise
• On the Robustness of Convolutional Neural Networks to Internal Architecture and Weight Perturbations
• Contrasting Adversarial Perturbations: The Space of Harmless Perturbations