mlsanityData to 14 Oct 2026

Disrupting Model Training with Adversarial Shortcuts

arXiv 2106.06654 ↗First posted 12 Jun 2021cs.CV cs.CR cs.LGcited by 7

Not on the map. Why?

Abstract (arXiv)

When data is publicly released for human consumption, it is unclear how to prevent its unauthorized usage for machine learning purposes. Successful model training may be preventable with carefully designed dataset modifications, and we present a proof-of-concept approach for the image classification setting. We propose methods based on the notion of adversarial shortcuts, which encourage models to rely on non-robust signals rather than semantic features, and our experiments demonstrate that these measures successfully prevent deep learning models from achieving high accuracy on real, unmodified data examples.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 25 of its 42 references that are in the map. First posted on arXiv 12 Jun 2021.

Builds on

25 of 42 references.

References · 25

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  5. 2016Densely Connected Convolutional Networkscited by 6,720
  6. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  7. 2016Understanding deep learning requires rethinking generalizationcited by 1,674
  8. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
  9. 2016SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model sizecited by 1,060
  10. 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 953
  11. 2018Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networkscited by 519
  12. 2018Confounding variables can degrade generalization performance of radiological deep learning modelscited by 379
13 more
  1. 2017Towards Poisoning of Deep Learning Algorithms with Back-gradient Optimizationcited by 197
  2. 2021Unlearnable Examples: Making Personal Data Unexploitablecited by 193
  3. 2020Witches' Brew: Industrial Scale Data Poisoning via Gradient Matchingcited by 163
  4. 2020MetaPoison: Practical General-purpose Clean-label Data Poisoningcited by 120
  5. 2020Just How Toxic is Data Poisoning? A Unified Benchmark for Backdoor and Data Poisoning Attackscited by 116
  6. 2020InstaHide: Instance-hiding Schemes for Private Distributed Learningcited by 90
  7. 2017Measuring the tendency of CNNs to Learn Surface Statistical Regularitiescited by 87
  8. 2021Manipulating SGD with Data Ordering Attackscited by 74
  9. 2020Bullseye Polytope: A Scalable Clean-Label Poisoning Attack with Improved Transferabilitycited by 66
  10. 2018Deep Learning Predicts Hip Fracture using Confounding Patient and Healthcare Variablescited by 50
  11. 2021Preventing Unauthorized Use of Proprietary Data: Poisoning for Secure Dataset Releasecited by 35
  12. 2019Poisoning Attacks with Generative Adversarial Netscited by 23
  13. 2020Is Private Learning Possible with Instance Encoding?cited by 22

Built on by

Cited by year

Citing papers per year: 2021: 2; 2022: 2; 2023: 2; 2024: 1.

7 papers.

  1. 2023Image Shortcut Squeezing: Countering Perturbative Availability Poisons with Compressioncited by 39
  2. 2022One-Pixel Shortcut: on the Learning Preference of Deep Neural Networkscited by 29
  3. 2024Provably Unlearnable Data Examplescited by 18
  4. 2021SoK: Anti-Facial Recognition Technologycited by 16
  5. 2021Fooling Adversarial Training with Inducing Noisecited by 13
  6. 2022Can Adversarial Training Be Manipulated By Non-Robust Features?cited by 12
  7. 2023On the Computational Entanglement of Distant Features in Adversarial Machine Learningcited by 0