mlsanityData to 7 Oct 2026

Detecting Adversarial Samples from Artifacts

arXiv 1703.00410 ↗First posted 1 Mar 2017stat.ML cs.LGcited by 244

Not on the map. Why?

Abstract (arXiv)

Deep neural networks (DNNs) are powerful nonlinear architectures that are known to be robust to random perturbations of the input. However, these models are vulnerable to adversarial perturbations--small input changes crafted explicitly to fool the model. In this paper, we ask whether a DNN can distinguish adversarial samples from their normal and noisy counterparts. We investigate model confidence on adversarial samples by looking at Bayesian uncertainty estimates, available in dropout neural networks, and by performing density estimation in the subspace of deep features learned by the model. The result is a method for implicit adversarial detection that is oblivious to the attack algorithm. We evaluate this method on a variety of standard datasets including MNIST and CIFAR-10 and show that it generalizes well across different architectures and attacks. Our findings report that 85-93% ROC-AUC can be achieved on a number of standard classification tasks with a negative class that consists of both normal and noisy samples.

Key facts. Cited by 244 papers in the map (data through 7 Oct 2026). First posted on arXiv 1 Mar 2017.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 76; 2022: 53; 2023: 50; 2024: 31; 2025: 22; 2026: 12.

244 papers.

Sort:

By area:

  1. 2023Glaze: Protecting Artists from Style Mimicry by Text-to-Image Modelscited by 200
  2. 2023A Survey of Safety and Trustworthiness of Large Language Models through the Lens of Verification and Validationcited by 91
  3. 2023X-Adv: Physical Adversarial Object Attacks against X-ray Prohibited Item Detectioncited by 58
  4. 2021Relating Adversarially Robust Generalization to Flat Minimacited by 57
  5. 2023Active Instruction Tuning: Improving Cross-Task Generalization by Training on Prompt Sensitive Taskscited by 53
  6. 2021Adversarial Example Detection for DNN Models: A Review and Experimental Comparisoncited by 51
  7. 2021Improving black-box optimization in VAE latent space using decoder uncertaintycited by 51
  8. 2021Knowing More About Questions Can Help: Improving Calibration in Question Answeringcited by 47
  9. 2022Patch-Fool: Are Vision Transformers Always Robust Against Adversarial Perturbations?cited by 47
  10. 2021Detecting Adversarial Examples Is (Nearly) As Hard As Classifying Themcited by 46
  11. 2022A Survey of Robust Adversarial Training in Pattern Recognition: Fundamental, Theory, and Methodologiescited by 44
  12. 2021Optical Adversarial Attackcited by 43
38 more
  1. 2021LiBRe: A Practical Bayesian Approach to Adversarial Detectioncited by 40
  2. 2021Biometrics: Trust, but Verifycited by 39
  3. 2021Inspect, Understand, Overcome: A Survey of Practical Methods for AI Safetycited by 35
  4. 2021Evading Adversarial Example Detection Defenses with Orthogonal Projected Gradient Descentcited by 34
  5. 2021A Little Robustness Goes a Long Way: Leveraging Robust Features for Targeted Transfer Attackscited by 33
  6. 2021SpectralDefense: Detecting Adversarial Attacks on CNNs in the Fourier Domaincited by 30
  7. 2021Low Curvature Activations Reduce Overfitting in Adversarial Trainingcited by 29
  8. 2021Detecting Adversarial Examples from Sensitivity Inconsistency of Spatial-Transform Domaincited by 29
  9. 2021Drawing Robust Scratch Tickets: Subnetworks with Inborn Robustness Are Found within Randomly Initialized Networkscited by 29
  10. 2021A Review of Adversarial Attack and Defense for Classification Methodscited by 29
  11. 2022Benchmarking Adversarially Robust Quantum Machine Learning at Scalecited by 28
  12. 2022iDECODe: In-distribution Equivariance for Conformal Out-of-distribution Detectioncited by 27
  13. 2022Towards Realistic Semi-Supervised Learningcited by 27
  14. 2021Adversarial Examples in Multi-Layer Random ReLU Networkscited by 26
  15. 2023Towards quantum enhanced adversarial robustness in machine learningcited by 26
  16. 2021A Survey on Open Set Recognitioncited by 25
  17. 2022Robust Models are less Over-Confidentcited by 25
  18. 2021Adversarial Attacks on Deep Models for Financial Transaction Recordscited by 24
  19. 2021ML Attack Models: Adversarial Attacks and Data Poisoning Attackscited by 23
  20. 2021Taxonomy of Machine Learning Safety: A Survey and Primercited by 22
  21. 2023Detecting Adversarial Data by Probing Multiple Perturbations Using Expected Perturbation Scorecited by 20
  22. 2021Adversarial Robustness via Fisher-Rao Regularizationcited by 19
  23. 2024MirrorCheck: Efficient Adversarial Defense for Vision-Language Modelscited by 19
  24. 2021SPADE: A Spectral Method for Black-Box Adversarial Robustness Evaluationcited by 16
  25. 2022Self-Ensembling Vision Transformer (SEViT) for Robust Medical Image Classificationcited by 16
  26. 2022LLMEffiChecker: Understanding and Testing Efficiency Degradation of Large Language Modelscited by 16
  27. 2024AcTracer: Active Testing of Large Language Model via Multi-Stage Samplingcited by 16
  28. 2021Towards Evaluating the Robustness of Deep Diagnostic Models by Adversarial Attackcited by 15
  29. 2021Explainable Deep Learning in Healthcare: A Methodological Survey from an Attribution Viewcited by 15
  30. 2024LLM Whisperer: An Inconspicuous Attack to Bias LLM Responsescited by 15
  31. 2021Attack as Defense: Characterizing Adversarial Examples using Robustnesscited by 14
  32. 2021Two Coupled Rejection Metrics Can Tell Adversarial Examples Apartcited by 14
  33. 2023Defenses in Adversarial Machine Learning: A Surveycited by 14
  34. 2021Interpreting Attributions and Interactions of Adversarial Attackscited by 13
  35. 2022Deep Learning Serves Traffic Safety Analysis: A Forward-looking Reviewcited by 13
  36. 2021Exploiting Multi-Object Relationships for Detecting Adversarial Attacks in Complex Scenescited by 12
  37. 2022On the Robustness of Bayesian Neural Networks to Adversarial Attackscited by 12
  38. 2021On the Robustness of Domain Constraintscited by 11

50 most cited of 244 All 244 (JSON)