mlsanityData to 7 Oct 2026

Temporal Shuffling for Defending Deep Action Recognition Models against Adversarial Attacks

arXiv 2112.07921 ↗First posted 15 Dec 2021cs.CVcited by 7

Not on the map. Why?

Abstract (arXiv)

Recently, video-based action recognition methods using convolutional neural networks (CNNs) achieve remarkable recognition performance. However, there is still lack of understanding about the generalization mechanism of action recognition models. In this paper, we suggest that action recognition models rely on the motion information less than expected, and thus they are robust to randomization of frame orders. Furthermore, we find that motion monotonicity remaining after randomization also contributes to such robustness. Based on this observation, we develop a novel defense method using temporal shuffling of input videos against adversarial attacks for action recognition models. Another observation enabling our defense method is that adversarial perturbations on videos are sensitive to temporal destruction. To the best of our knowledge, this is the first attempt to design a defense method without additional training for 3D CNN-based video action recognition models.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Builds on 31 of its 41 references that are in the map. First posted on arXiv 15 Dec 2021.

Builds on

31 of 41 references.

References · 31

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2017Quo Vadis, Action Recognition? A New Model and the Kinetics Datasetcited by 3,259
  4. 2017Non-local Neural Networkscited by 2,290
  5. 2017The Kinetics Human Action Video Datasetcited by 1,894
  6. 2018SlowFast Networks for Video Recognitioncited by 1,862
  7. 2016Adversarial examples in the physical worldcited by 1,762
  8. 2021Is Space-Time Attention All You Need for Video Understanding?cited by 1,609
  9. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  10. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
  11. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,259
  12. 2018Recognition in Terra Incognitacited by 655
19 more
  1. 2020X3D: Expanding Architectures for Efficient Video Recognitioncited by 635
  2. 2020On Adaptive Attacks to Adversarial Example Defensescited by 547
  3. 2014Long-term Recurrent Convolutional Networks for Visual Recognition and Descriptioncited by 483
  4. 2017AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actionscited by 449
  5. 2017Mitigating Adversarial Effects Through Randomizationcited by 382
  6. 2020Noise or Signal: The Role of Image Backgrounds in Object Recognitioncited by 342
  7. 2019High Frequency Component Helps Explain the Generalization of Convolutional Neural Networkscited by 310
  8. 2019Video Classification with Channel-Separated Convolutional Networkscited by 290
  9. 2016A study of the effect of JPG compression on adversarial imagescited by 196
  10. 2017Towards Robust Neural Networks via Random Self-ensemblecited by 158
  11. 2017Mitigating Evasion Attacks to Deep Neural Networks via Region-based Classificationcited by 56
  12. 2018Sparse Adversarial Perturbations for Videoscited by 56
  13. 2018Not Using the Car to See the Sidewalk: Quantifying and Controlling the Effects of Context in Classification and Segmentationcited by 49
  14. 2016Object Recognition with and without Objectscited by 39
  15. 2020Over-the-Air Adversarial Flickering Attacks against Video Recognition Networkscited by 35
  16. 2018Adversarial Perturbations Against Real-Time Video Classification Systemscited by 32
  17. 2021Action Transformer: A Self-Attention Model for Short-Time Pose-Based Human Action Recognitioncited by 24
  18. 2020Defending Against Multiple and Unforeseen Adversarial Videoscited by 19
  19. 2020Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attackcited by 12

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 3; 2025: 3; 2026: 1.

7 papers.

  1. 2024Multimodal Attack Detection for Action Recognition Modelscited by 5
  2. 2024Query-Efficient Video Adversarial Attack with Stylized Logo on Service Computingcited by 3
  3. 2024Detecting Adversarial Examplescited by 2
  4. 2025Universal and Efficient Detection of Adversarial Data through Nonuniform Impact on Network Layerscited by 2
  5. 2025Adversarial Video Promotion Against Text-to-Video Retrievalcited by 1
  6. 2025FeatureFool: Zero-Query Fooling of Video Models via Feature Mapcited by 0
  7. 2026FMVP: Masked Flow Matching for Adversarial Video Purificationcited by 0