mlsanityData to 7 Oct 2026

Deflecting Adversarial Attacks with Pixel Deflection

arXiv 1801.08926 ↗First posted 26 Jan 2018cs.CV cs.CRcited by 71

Not on the map. Why?

Abstract (arXiv)

CNNs are poised to become integral parts of many critical systems. Despite their robustness to natural variations, image pixel values can be manipulated, via small, carefully crafted, imperceptible perturbations, to cause a model to misclassify images. We present an algorithm to process an image so that classification accuracy is significantly preserved in the presence of such adversarial manipulations. Image classifiers tend to be robust to natural noise, and adversarial attacks tend to be agnostic to object location. These observations motivate our strategy, which leverages model robustness to defend against adversarial perturbations by forcing the image to match natural image statistics. Our algorithm locally corrupts the image by redistributing pixel values via a process we term pixel deflection. A subsequent wavelet-based denoising operation softens this corruption, as well as some of the adversarial changes. We demonstrate experimentally that the combination of these techniques enables the effective recovery of the true class, against a variety of robust attacks. Our results compare favorably with current state-of-the-art defenses, without requiring retraining or modifying the CNN.

Key facts. Cited by 71 papers in the map (data through 7 Oct 2026). First posted on arXiv 26 Jan 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 22; 2022: 17; 2023: 7; 2024: 11; 2025: 10; 2026: 4.

71 papers.

Sort:

By area:

  1. 2021AdvDrop: Adversarial Attack to DNNs by Dropping Informationcited by 55
  2. 2021Adversarial Example Detection for DNN Models: A Review and Experimental Comparisoncited by 51
  3. 2022DISCO: Adversarial Defense with Local Implicit Functionscited by 33
  4. 2022Fooling the Eyes of Autonomous Vehicles: Robust Physical Adversarial Examples Against Traffic Sign Recognition Systemscited by 27
  5. 2023RFLA: A Stealthy Reflected Light Adversarial Attack in the Physical Worldcited by 23
  6. 2022Adversarial Attack on Attackers: Post-Process to Mitigate Black-Box Score-Based Query Attackscited by 18
  7. 2021MagDR: Mask-guided Detection and Reconstruction for Defending Deepfakescited by 17
  8. 2021Attack to Fool and Explain Deep Networkscited by 17
  9. 2021Stereoscopic Universal Perturbations across Different Architectures and Datasetscited by 17
  10. 2021Detecting Adversarial Examples by Input Transformations, Defense Perturbations, and Votingcited by 16
  11. 2022Towards Good Practices in Evaluating Transfer Adversarial Attackscited by 14
  12. 2021Local Competition and Stochasticity for Adversarial Robustness in Deep Learningcited by 12
38 more
  1. 2022Rethinking Bayesian Learning for Data Analysis: The Art of Prior and Inference in Sparsity-Aware Modelingcited by 12
  2. 2022Understanding Adversarial Robustness Against On-manifold Adversarial Examplescited by 12
  3. 2023Boosting Black-box Attack to Deep Neural Networks with Conditional Diffusion Modelscited by 12
  4. 2022Robust Real-World Image Super-Resolution against Adversarial Attackscited by 11
  5. 2025Vanish into Thin Air: Cross-prompt Universal Adversarial Attacks for SAM2cited by 11
  6. 2022Enhancing Adversarial Robustness for Deep Metric Learningcited by 10
  7. 2025Towards Model Resistant to Transferable Adversarial Examples via Trigger Activationcited by 10
  8. 2021Revisiting Model's Uncertainty and Confidences for Adversarial Example Detectioncited by 9
  9. 2023Sub-token ViT Embedding via Stochastic Resonance Transformerscited by 9
  10. 2021Adversarial Attack and Defense in Deep Rankingcited by 8
  11. 2024Improving Black-box Robustness with In-Context Rewritingcited by 8
  12. 2022Self-recoverable Adversarial Examples: A New Effective Protection Mechanism in Social Networkscited by 6
  13. 2022How to Fine-tune Models with Few Samples: Update, Data Augmentation, and Test-time Augmentationcited by 6
  14. 2022Improved Text Classification via Test-Time Augmentationcited by 6
  15. 2022Synergy with Translation Artifacts for Training and Inference in Multilingual Taskscited by 6
  16. 2023Frequency Domain Adversarial Training for Robust Volumetric Medical Segmentationcited by 6
  17. 2021All You Need is RAW: Defending Against Adversarial Attacks with Camera Image Pipelinescited by 5
  18. 2021Moir'e Attack (MA): A New Potential Risk of Screen Photoscited by 4
  19. 2022CSTAR: Towards Compact and STructured Deep Neural Networks with Adversarial Robustnesscited by 4
  20. 2022General Adversarial Defense Against Black-box Attacks via Pixel Level and Feature Level Distribution Alignmentscited by 4
  21. 2024Mitigating Adversarial Perturbations for Deep Reinforcement Learning via Vector Quantizationcited by 4
  22. 2025ETA: Energy-based Test-time Adaptation for Depth Completioncited by 4
  23. 2022Treatment Learning Causal Transformer for Noisy Image Classificationcited by 3
  24. 2023PEARL: Preprocessing Enhanced Adversarial Robust Learning of Image Deraining for Semantic Segmentationcited by 3
  25. 2024Enhancing the "Immunity" of Mixture-of-Experts Networks for Adversarial Defensecited by 3
  26. 2024Persistent Classification: A New Approach to Stability of Data and Adversarial Examplescited by 3
  27. 2021Sparta: Spatially Attentive and Adversarially Robust Activationcited by 2
  28. 2021Ensemble Defense with Data Diversity: Weak Correlation Implies Strong Robustnesscited by 2
  29. 2021Super-Efficient Super Resolution for Fast Adversarial Defense at the Edgecited by 2
  30. 2022Data-free Defense of Black Box Models Against Adversarial Attackscited by 2
  31. 2024Detecting Adversarial Examplescited by 2
  32. 2024CausAdv: A Causal-based Framework for Detecting Adversarial Examplescited by 2
  33. 2025Universal and Efficient Detection of Adversarial Data through Nonuniform Impact on Network Layerscited by 2
  34. 2021Adaptive Clustering of Robust Semantic Representations for Adversarial Image Purificationcited by 1
  35. 2021Generalized moduli of continuity under irregular or random deformations via multiscale analysiscited by 1
  36. 2021Robust Upper Bounds for Adversarial Trainingcited by 1
  37. 2022Adversarial samples for deep monocular 6D object pose estimationcited by 1
  38. 2024ARQ: A Mixed-Precision Quantization Framework for Accurate and Certifiably Robust DNNscited by 1

50 most cited of 71 All 71 (JSON)