mlsanityData to 14 Oct 2026

Neural Polarizer: A Lightweight and Effective Backdoor Defense via Purifying Poisoned Features

arXiv 2306.16697 ↗First posted 29 Jun 2023cs.AI cs.CRcited by 37

Not on the map. Why?

Abstract (arXiv)

Recent studies have demonstrated the susceptibility of deep neural networks to backdoor attacks. Given a backdoored model, its prediction of a poisoned sample with trigger will be dominated by the trigger information, though trigger information and benign information coexist. Inspired by the mechanism of the optical polarizer that a polarizer could pass light waves with particular polarizations while filtering light waves with other polarizations, we propose a novel backdoor defense method by inserting a learnable neural polarizer into the backdoored model as an intermediate layer, in order to purify the poisoned sample via filtering trigger information while maintaining benign information. The neural polarizer is instantiated as one lightweight linear transformation layer, which is learned through solving a well designed bi-level optimization problem, based on a limited clean dataset. Compared to other fine-tuning-based defense methods which often adjust all parameters of the backdoored model, the proposed method only needs to learn one additional layer, such that it is more efficient and requires less clean data. Extensive experiments demonstrate the effectiveness and efficiency of our method in removing backdoors across various neural network architectures and datasets, especially in the case of very limited clean data.

Key facts. Cited by 37 papers in the map (data through 14 Oct 2026). Builds on 36 of its 56 references that are in the map. First posted on arXiv 29 Jun 2023.

Builds on

36 of 56 references.

References · 36

  1. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  3. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  4. 2016Adversarial examples in the physical worldcited by 1,769
  5. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,318
  6. 2017Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoningcited by 1,055
  7. 2018Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networkscited by 580
  8. 2018Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networkscited by 519
  9. 2018Black-box Adversarial Attacks with Limited Queries and Informationcited by 501
  10. 2020Invisible Backdoor Attack with Sample-Specific Triggerscited by 353
  11. 2017Defense against Adversarial Attacks Using High-Level Representation Guided Denoisercited by 329
  12. 2021Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networkscited by 324
24 more
  1. 2018Feature Denoising for Improving Adversarial Robustnesscited by 314
  2. 2020Input-Aware Dynamic Backdoor Attackcited by 313
  3. 2021Recent Advances in Adversarial Training for Adversarial Robustnesscited by 279
  4. 2019A new Backdoor Attack in CNNs by training set corruption without label poisoningcited by 242
  5. 2021Adversarial Neuron Pruning Purifies Backdoored Deep Modelscited by 235
  6. 2020Understanding and Improving Fast Adversarial Trainingcited by 214
  7. 2021Adversarial Unlearning of Backdoors via Implicit Hypergradientcited by 166
  8. 2021Rethinking the Backdoor Attacks' Triggers: A Frequency Perspectivecited by 161
  9. 2022Backdoor Defense via Decoupling the Training Processcited by 153
  10. 2022BackdoorBench: A Comprehensive Benchmark of Backdoor Learningcited by 128
  11. 2018Transferable Adversarial Attacks for Image and Video Object Detectioncited by 109
  12. 2022Parallel Rectangle Flip Attack: A Query-based Black-box Attack against Object Detectioncited by 79
  13. 2022Data-free Backdoor Removal based on Channel Lipschitznesscited by 79
  14. 2020Efficient Adversarial Attacks for Visual Object Trackingcited by 66
  15. 2023Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimizationcited by 64
  16. 2022The "Beatrix'' Resurrections: Robust Backdoor Detection via Gram Matricescited by 63
  17. 2021AEVA: Black-box Backdoor Detection Using Adversarial Extreme Value Analysiscited by 53
  18. 2023Backdoor Defense via Adaptively Splitting Poisoned Datasetcited by 43
  19. 2020Open-set Adversarial Defensecited by 27
  20. 2022Imperceptible and Robust Backdoor Attack in 3D Point Cloudcited by 25
  21. 2022Hardly Perceptible Trojan Attack against Neural Networks with Bit Flipscited by 22
  22. 2022Open-set Adversarial Defense with Clean-Adversarial Mutual Learningcited by 21
  23. 2021Exploring Targeted Universal Adversarial Perturbations to End-to-end ASR Modelscited by 10
  24. 2023Mean Parity Fair Regression in RKHScited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 5; 2024: 18; 2025: 8; 2026: 6.

37 papers.

Sort:

By area:

  1. 2023BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learningcited by 115
  2. 2023BadCLIP: Trigger-Aware Prompt Learning for Backdoor Attacks on CLIPcited by 67
  3. 2023Shared Adversarial Unlearning: Backdoor Mitigation by Unlearning Shared Adversarial Examplescited by 43
  4. 2024Breaking the False Sense of Security in Backdoor Defense through Re-Activation Attackcited by 28
  5. 2025REFINE: Inversion-Free Backdoor Defense via Model Reprogrammingcited by 22
  6. 2024BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learningcited by 19
  7. 2023Defenses in Adversarial Machine Learning: A Surveycited by 14
  8. 2024CleanerCLIP: Fine-grained Counterfactual Semantic Augmentation for Backdoor Defense in Contrastive Learningcited by 13
  9. 2024Defending Multimodal Backdoored Models by Repulsive Visual Prompt Tuningcited by 12
  10. 2024Mitigating Backdoor Attack by Injecting Proactive Defensive Backdoorcited by 11
  11. 2024BAN: Detecting Backdoors Activated by Adversarial Neuron Noisecited by 8
  12. 2024Mutual Information Guided Backdoor Mitigation for Pre-trained Encoderscited by 8
25 more
  1. 2025Bridging Symmetry and Robustness: On the Role of Equivariance in Enhancing Adversarial Robustnesscited by 8
  2. 2025Breaking the Stealth-Potency Trade-off in Clean-Image Backdoors with Generative Trigger Optimizationcited by 7
  3. 2024TrojFM: Resource-efficient Backdoor Attacks against Very Large Foundation Modelscited by 5
  4. 2025TokenSwap: Backdoor Attack on the Compositional Understanding of Large Vision-Language Modelscited by 5
  5. 2026Test-Time Attention Purification for Backdoored Large Vision Language Modelscited by 5
  6. 2024Perturb and Recover: Fine-tuning for Effective Backdoor Removal from CLIPcited by 4
  7. 2024Event Trojan: Asynchronous Event-based Backdoor Attackscited by 3
  8. 2024Reliable Poisoned Sample Detection against Backdoor Attacks Enhanced by Sharpness Aware Minimizationcited by 3
  9. 2025ReVeil: Unconstrained Concealed Backdoor Attack on Deep Neural Networks using Machine Unlearningcited by 3
  10. 2026Explainability-Guided Defense: Attribution-Aware Model Refinement Against Adversarial Data Attackscited by 3
  11. 2023Erasing Self-Supervised Learning Backdoor by Cluster Activation Maskingcited by 2
  12. 2024Backdoor Defense through Self-Supervised and Generative Learningcited by 2
  13. 2024Combining Machine Learning Defenses without Conflictscited by 2
  14. 2024BadVim: Unveiling Backdoor Threats in Visual State Space Modelcited by 1
  15. 2024Prompting the Unseen: Detecting Hidden Backdoors in Black-Box Modelscited by 1
  16. 2024Trojan Cleansing with Neural Collapsecited by 1
  17. 2025Class-Conditional Neural Polarizer: A Lightweight and Effective Backdoor Defense by Purifying Poisoned Featurescited by 1
  18. 2025Neural Antidote: Class-Wise Prompt Tuning for Purifying Backdoors in CLIPcited by 1
  19. 2025AutoDetect: Designing an Autoencoder-based Detection Method for Poisoning Attacks on Object Detection Applications in the Military Domaincited by 1
  20. 2026Kill it with FIRE: On Leveraging Latent Space Directions for Runtime Backdoor Mitigation in Deep Neural Networkscited by 1
  21. 2024BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learningcited by 0
  22. 2024A general approach to enhance the survivability of backdoor attacks by decision path couplingcited by 0
  23. 2026PASTA: A Patch-Agnostic Twofold-Stealthy Backdoor Attack on Vision Transformerscited by 0
  24. 2026Rethinking Backdoor Repair Evaluation: Distinguishing Aggregate Clean Utility from Benign Performance Preservationcited by 0
  25. 2026TARE: Weigh a Never-Poisoned Twin Before Reading Backdoor-Defense Costscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023Beating Backdoor Attack at Its Own Game
  2. 2023Towards Stable Backdoor Purification through Feature Shift Tuning
  3. 2025The Eminence in Shadow: Exploiting Feature Boundary Ambiguity for Robust Backdoor Attacks
  4. 2024NBA: defensive distillation for backdoor removal via neural behavior alignment
  5. 2022Training with More Confidence: Mitigating Injected and Natural Backdoors During Training
  6. 2023A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives
  7. 2022Just Rotate it: Deploying Backdoor Attacks via Rotation Transformation
  8. 2022Defense against Backdoor Attacks via Identifying and Purifying Bad Neurons