mlsanityData to 14 Oct 2026

Adversarial and Clean Data Are Not Twins

arXiv 1704.04960 ↗First posted 17 Apr 2017cs.LG cs.NEcited by 52

Not on the map. Why?

Abstract (arXiv)

Adversarial attack has cast a shadow on the massive success of deep neural networks. Despite being almost visually identical to the clean data, the adversarial images can fool deep neural networks into wrong predictions with very high confidence. In this paper, however, we show that we can build a simple binary classifier separating the adversarial apart from the clean data with accuracy over 99%. We also empirically show that the binary classifier is robust to a second-round adversarial attack. In other words, it is difficult to disguise adversarial samples to bypass the binary classifier. Further more, we empirically investigate the generalization limitation which lingers on all current defensive methods, including the binary classifier approach. And we hypothesize that this is the result of intrinsic property of adversarial crafting algorithms.

Key facts. Cited by 52 papers in the map (data through 14 Oct 2026). First posted on arXiv 17 Apr 2017.

Builds on

No reference list yet.

Built on by

Cited by year

Citing papers per year: 2021: 26; 2022: 12; 2023: 10; 2024: 4.

52 papers.

Sort:

By area:

  1. 2023Certifying LLM Safety against Adversarial Promptingcited by 266
  2. 2021Trustworthy AI: A Computational Perspectivecited by 93
  3. 2021Adversarial Example Detection for DNN Models: A Review and Experimental Comparisoncited by 52
  4. 2022Practical Evaluation of Adversarial Robustness via Adaptive Auto Attackcited by 44
  5. 2021LiBRe: A Practical Bayesian Approach to Adversarial Detectioncited by 40
  6. 2021Policy Smoothing for Provably Robust Reinforcement Learningcited by 40
  7. 2021Biometrics: Trust, but Verifycited by 39
  8. 2022How to Robustify Black-Box ML Models? A Zeroth-Order Optimization Perspectivecited by 39
  9. 2021Unified Detection of Digital and Physical Face Attackscited by 31
  10. 2021A Review of Adversarial Attack and Defense for Classification Methodscited by 29
  11. 2021Taxonomy of Machine Learning Safety: A Survey and Primercited by 25
  12. 2022Defensive Patches for Robust Recognition in the Physical Worldcited by 20
38 more
  1. 2021Universal Adversarial Perturbations Through the Lens of Deep Steganography: Towards A Fourier Perspectivecited by 18
  2. 2021Center Smoothing: Certified Robustness for Networks with Structured Outputscited by 18
  3. 2021Two Coupled Rejection Metrics Can Tell Adversarial Examples Apartcited by 14
  4. 2021PatchCensor: Patch Robustness Certification for Transformers via Exhaustive Testingcited by 12
  5. 2022Visual Adversarial Attacks and Defenses in the Physical World: A Surveycited by 11
  6. 2021AID-Purifier: A Light Auxiliary Network for Boosting Adversarial Defensecited by 10
  7. 2021Improving Model Robustness with Latent Distribution Locally and Globallycited by 9
  8. 2022Self-recoverable Adversarial Examples: A New Effective Protection Mechanism in Social Networkscited by 6
  9. 2023Investigating Catastrophic Overfitting in Fast Adversarial Training: A Self-fitting Perspectivecited by 5
  10. 2021Explaining Adversarial Vulnerability with a Data Sparsity Hypothesiscited by 4
  11. 2021Attribution of Gradient Based Adversarial Attacks for Reverse Engineering of Deceptionscited by 4
  12. 2021Long-term Cross Adversarial Training: A Robust Meta-learning Method for Few-shot Classification Taskscited by 4
  13. 2022Resisting Adversarial Attacks in Deep Neural Networks using Diverse Decision Boundariescited by 4
  14. 2024Fortify the Guardian, Not the Treasure: Resilient Adversarial Detectorscited by 4
  15. 2021ExAD: An Ensemble Approach for Explanation-based Adversarial Detectioncited by 3
  16. 2023AdvFAS: A robust face anti-spoofing framework against adversarial examplescited by 3
  17. 2021Noise Sensitivity-Based Energy Efficient and Robust Adversary Detection in Neural Networkscited by 2
  18. 2021Adversarial Examples Detection with Bayesian Neural Networkcited by 2
  19. 2021Two Souls in an Adversarial Image: Towards Universal Adversarial Example Detection using Multi-view Inconsistencycited by 2
  20. 2022Adversarial Detection without Model Informationcited by 2
  21. 2024Rethinking the Threat and Accessibility of Adversarial Attacks against Face Recognition Systemscited by 2
  22. 2022White-Box Attacks on Hate-speech BERT Classifiers in German with Explicit and Implicit Character Level Defensecited by 1
  23. 2023Provable Robustness for Streaming Models with a Sliding Windowcited by 1
  24. 2024HOLMES: to Detect Adversarial Examples with Multiple Detectorscited by 1
  25. 2021PredCoin: Defense against Query-based Hard-label Attackcited by 0
  26. 2021Using Anomaly Feature Vectors for Detecting, Classifying and Warning of Outlier Adversarial Examplescited by 0
  27. 2021Unsupervised Detection of Adversarial Examples with Model Explanationscited by 0
  28. 2021Models of Computational Profiles to Study the Likelihood of DNN Metamorphic Test Casescited by 0
  29. 2021Detecting Adversaries, yet Faltering to Noise? Leveraging Conditional Variational AutoEncoders for Adversary Detection in the Presence of Noisy Imagescited by 0
  30. 2022Hindi/Bengali Sentiment Analysis Using Transfer Learning and Joint Dual Input Learning with Self Attentioncited by 0
  31. 2022HoneyModels: Machine Learning Honeypotscited by 0
  32. 2022Btech thesis report on adversarial attack detection and purification of adverserially attacked imagescited by 0
  33. 2022Adversarial Example Detection in Deployed Tree Ensemblescited by 0
  34. 2023Mobile Edge Adversarial Detection for Digital Twinning to the Metaverse with Deep Reinforcement Learningcited by 0
  35. 2023Unsupervised Multi-Criteria Adversarial Detection in Deep Image Retrievalcited by 0
  36. 2023Adaptive Attractors: A Defense Strategy against ML Adversarial Collusion Attackscited by 0
  37. 2023Adversarial Attacks Neutralization via Data Set Randomizationcited by 0
  38. 2023SysNoise: Exploring and Benchmarking Training-Deployment System Inconsistencycited by 0

50 most cited of 52 All 52 (JSON)