mlsanityData to 14 Oct 2026

Defending against Patch-Based and Texture-Based Adversarial Attacks with Spectral Decomposition

arXiv 2604.10715 ↗First posted 12 Apr 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Adversarial examples present significant challenges to the security of Deep Neural Network (DNN) applications. Specifically, there are patch-based and texture-based attacks that are usually used to craft physical-world adversarial examples, posing real threats to security-critical applications such as person detection in surveillance and autonomous systems, because those attacks are physically realizable. Existing defense mechanisms face challenges in the adaptive attack setting, i.e., the attacks are specifically designed against them. In this paper, we propose Adversarial Spectrum Defense (ASD), a defense mechanism that leverages spectral decomposition via Discrete Wavelet Transform (DWT) to analyze adversarial patterns across multiple frequency scales. The multi-resolution and localization capability of DWT enables ASD to capture both high-frequency (fine-grained) and low-frequency (spatially pervasive) perturbations. By integrating this spectral analysis with the off-the-shelf Adversarial Training (AT) model, ASD provides a comprehensive defense strategy against both patch-based and texture-based adversarial attacks. Extensive experiments demonstrate that ASD+AT achieved state-of-the-art (SOTA) performance against various attacks, outperforming the APs of previous defense methods by 21.73%, in the face of strong adaptive adversaries specifically designed against ASD. Code available at https://github.com/weiz0823/adv-spectral-defense .

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 32 of its 50 references that are in the map. First posted on arXiv 12 Apr 2026.

Builds on

32 of 50 references.

References · 32

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  5. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  6. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  7. 2016Towards Evaluating the Robustness of Neural Networkscited by 2,998
  8. 2016Adversarial examples in the physical worldcited by 1,769
  9. 2019FCOS: Fully Convolutional One-Stage Object Detectioncited by 1,577
  10. 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,451
  11. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
  12. 2020On Adaptive Attacks to Adversarial Example Defensescited by 555
20 more
  1. 2017Adversarial Patchcited by 517
  2. 2019Fooling automated surveillance cameras: adversarial patches to attack person detectioncited by 268
  3. 2019Making an Invisibility Cloak: Real World Adversarial Attacks on Object Detectorscited by 155
  4. 2018Feature Distillation: DNN-Oriented JPEG Compression Against Adversarial Examplescited by 99
  5. 2022Adversarial Texture for Fooling Person Detectors in the Physical Worldcited by 96
  6. 2018Local Gradients Smoothing: Defense against localized adversarial attackscited by 87
  7. 2025RF-DETR: Neural Architecture Search for Real-Time Detection Transformerscited by 75
  8. 2021Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domaincited by 71
  9. 2021Segment and Complete: Defending Object Detectors against Adversarial Patch Attacks with Robust Patch Detectioncited by 67
  10. 2022Physical Adversarial Attack meets Computer Vision: A Decade Surveycited by 67
  11. 2023Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modelingcited by 41
  12. 2023Jedi: Entropy-based Localization and Removal of Adversarial Patchescited by 35
  13. 2024PAD: Patch-Agnostic Defense against Adversarial Patch Attackscited by 32
  14. 2022PatchZero: Defending against Adversarial Patch Attacks by Detecting and Zeroing the Patchcited by 30
  15. 2021Adversarial YOLO: Defense Human Detection Patch Attacks via Detecting Adversarial Patchescited by 17
  16. 2024NumbOD: A Spatial-Frequency Fusion Attack Against Object Detectorscited by 17
  17. 2023On the Importance of Backbone to the Adversarial Robustness of Object Detectorscited by 12
  18. 2023Improving Adversarial Robustness of Masked Autoencoders via Test-time Frequency-domain Promptingcited by 12
  19. 2022LPF-Defense: 3D Adversarial Defense based on Frequency Analysiscited by 10
  20. 2022Defending From Physically-Realizable Adversarial Attacks Through Internal Over-Activation Analysiscited by 10

Built on by

Not cited yet.