On the Interplay of Convolutional Padding and Adversarial Robustness
Not on the map. Why?
Abstract (arXiv)
It is common practice to apply padding prior to convolution operations to preserve the resolution of feature-maps in Convolutional Neural Networks (CNN). While many alternatives exist, this is often achieved by adding a border of zeros around the inputs. In this work, we show that adversarial attacks often result in perturbation anomalies at the image boundaries, which are the areas where padding is used. Consequently, we aim to provide an analysis of the interplay between padding and adversarial attacks and seek an answer to the question of how different padding modes (or their absence) affect adversarial robustness in various scenarios.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 45 of its 66 references that are in the map. First posted on arXiv 12 Aug 2023.
Builds on
45 of 66 references.
References · 45
- 2015Deep Residual Learning for Image Recognitioncited by 52,096
- 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
- 2014Explaining and Harnessing Adversarial Examplescited by 7,364
- 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
- 2013Intriguing properties of neural networkscited by 5,155
- 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 5,045
- 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,451
- 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
- 2016Adversarial Machine Learning at Scalecited by 936
- 2020Fast is better than free: Revisiting adversarial trainingcited by 727
- 2020Inverting Gradients -- How easy is it to break privacy in federated learning?cited by 712
- 2017Evasion Attacks against Machine Learning at Test Timecited by 689
33 more
- 2019Square Attack: a query-efficient black-box adversarial attack via random searchcited by 685
- 2017One pixel attack for fooling deep neural networkscited by 582
- 2016Delving into Transferable Adversarial Examples and Black-box Attackscited by 536
- 2020Overfitting in adversarially robust deep learningcited by 506
- 2018Black-box Adversarial Attacks with Limited Queries and Informationcited by 501
- 2018Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Surveycited by 460
- 2019On Evaluating Adversarial Robustnesscited by 447
- 2019Minimally distorted Adversarial Examples with a Fast Adaptive Boundary Attackcited by 334
- 2020Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributionscited by 280
- 2020Uncovering the Limits of Adversarial Training against Norm-Bounded Adversarial Examplescited by 278
- 2018Towards Fast Computation of Certified Robustness for ReLU Networkscited by 255
- 2021Improving Robustness using Generated Datacited by 234
- 2022MAT: Mask-Aware Transformer for Large Hole Image Inpaintingcited by 218
- 2020Understanding and Improving Fast Adversarial Trainingcited by 214
- 2021Data Augmentation Can Improve Robustnesscited by 194
- 2020How Much Position Information Do Convolutional Neural Networks Encode?cited by 172
- 2019Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Drivingcited by 152
- 2020On Translation Invariance in CNNs: Convolutional Layers can Exploit Absolute Spatial Locationcited by 142
- 2021Stable Neural ODE with Lyapunov-Stable Equilibrium Points for Defending Against Adversarial Attackscited by 91
- 2021RobustART: Benchmarking Robustness on Architecture Design and Training Techniquescited by 79
- 2020Understanding Catastrophic Overfitting in Single-step Adversarial Trainingcited by 76
- 2016Learning to Refine Object Segmentscited by 67
- 2018Training for Faster Adversarial Robustness Verification via Inducing ReLU Stabilitycited by 67
- 2019InteractE: Improving Convolution-based Knowledge Graph Embeddings by Increasing Feature Interactionscited by 58
- 2020Mind the Pad -- CNNs can Develop Blind Spotscited by 41
- 2022FrequencyLowCut Pooling -- Plug & Play against Catastrophic Overfittingcited by 28
- 2022Revisiting Residual Networks for Adversarial Robustness: An Architectural Perspectivecited by 28
- 2022Robust Models are less Over-Confidentcited by 25
- 2022CNN Filter DB: An Empirical Investigation of Trained Convolutional Filterscited by 20
- 2023Neural Architecture Design and Robustness: A Datasetcited by 15
- 2021Spectral Leakage and Rethinking the Kernel Size in CNNscited by 11
- 2022Adversarial Robustness through the Lens of Convolutional Filterscited by 9
- 2023An Extended Study of Human-like Behavior under Adversarial Trainingcited by 9
Built on by
Citing papers per year: 2024: 1.
1 paper.