mlsanityData to 14 Oct 2026

Model Agnostic Defence against Backdoor Attacks in Machine Learning

arXiv 1908.02203 ↗First posted 6 Aug 2019cs.LG cs.CRcited by 52

Not on the map. Why?

Abstract (arXiv)

Machine Learning (ML) has automated a multitude of our day-to-day decision making domains such as education, employment and driving automation. The continued success of ML largely depends on our ability to trust the model we are using. Recently, a new class of attacks called Backdoor Attacks have been developed. These attacks undermine the user's trust in ML models. In this work, we present NEO, a model agnostic framework to detect and mitigate such backdoor attacks in image classification ML models. For a given image classification model, our approach analyses the inputs it receives and determines if the model is backdoored. In addition to this feature, we also mitigate these attacks by determining the correct predictions of the poisoned images. An appealing feature of NEO is that it can, for the first time, isolate and reconstruct the backdoor trigger. NEO is also the first defence methodology, to the best of our knowledge that is completely blackbox. We have implemented NEO and evaluated it against three state of the art poisoned models. These models include highly critical applications such as traffic sign detection (USTS) and facial detection. In our evaluation, we show that NEO can detect ≈88% of the poisoned inputs on average and it is as fast as 4.4 ms per input image. We also reconstruct the poisoned input for the user to effectively test their systems.

Key facts. Cited by 52 papers in the map (data through 14 Oct 2026). First posted on arXiv 6 Aug 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 7; 2022: 14; 2023: 11; 2024: 12; 2025: 6; 2026: 2.

52 papers.

Sort:

By area:

  1. 2022BackdoorBench: A Comprehensive Benchmark of Backdoor Learningcited by 128
  2. 2022Measuring the Effect of Training Data on Deep Learning Predictions via Randomized Experimentscited by 50
  3. 2023Detecting Backdoors During the Inference Stage Based on Corruption Robustness Consistencycited by 46
  4. 2023Black-box Backdoor Defense via Zero-shot Image Purificationcited by 45
  5. 2022Towards A Proactive ML Approach for Detecting Backdoor Poison Samplescited by 40
  6. 2022Fine-Tuning Is All You Need to Mitigate Backdoor Attackscited by 38
  7. 2021Towards Practical Deployment-Stage Backdoor Attack on Deep Neural Networkscited by 34
  8. 2021De-Pois: An Attack-Agnostic Defense against Data Poisoning Attackscited by 33
  9. 2022Identifying a Training-Set Attack's Target Using Renormalized Influence Estimationcited by 32
  10. 2022Defending Backdoor Attacks on Vision Transformer via Patch Processingcited by 20
  11. 2024BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learningcited by 19
  12. 2021Textual Backdoor Attacks Can Be More Harmful via Two Simple Trickscited by 18
38 more
  1. 2021Backdoor Attack through Frequency Domaincited by 18
  2. 2022A Survey of Neural Trojan Attacks and Defenses in Deep Learningcited by 17
  3. 2022Quarantine: Sparsity Can Uncover the Trojan Attack Trigger for Freecited by 15
  4. 2024Preference Poisoning Attacks on Reward Model Learningcited by 14
  5. 2025Test-Time Backdoor Detection for Object Detection Modelscited by 14
  6. 2021WaNet -- Imperceptible Warping-based Backdoor Attackcited by 5
  7. 2023Influencer Backdoor Attack on Semantic Segmentationcited by 5
  8. 2024Test-Time Multimodal Backdoor Detection by Contrastive Promptingcited by 5
  9. 2025DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Datacited by 5
  10. 2021MISA: Online Defense of Trojaned Models using Misattributionscited by 4
  11. 2023Towards Understanding How Self-training Tolerates Data Backdoor Poisoningcited by 4
  12. 2023SecurityNet: Assessing Machine Learning Vulnerabilities on Public Modelscited by 4
  13. 2022Trojan Horse Training for Breaking Defenses against Backdoor Attacks in Deep Learningcited by 3
  14. 2022Emerging Threats in Deep Learning-Based Autonomous Driving: A Comprehensive Surveycited by 3
  15. 2023INK: Inheritable Natural Backdoor Attack Against Model Distillationcited by 3
  16. 2023Adversary Aware Continual Learningcited by 3
  17. 2024The Art of Deception: Robust Backdoor Attack using Dynamic Stacking of Triggerscited by 3
  18. 2024BackdoorMBTI: A Backdoor Learning Multimodal Benchmark Tool Kit for Backdoor Defense Evaluationcited by 3
  19. 2022Learned-Database Systems Securitycited by 2
  20. 2024On the (In)feasibility of ML Backdoor Detection as an Hypothesis Testing Problemcited by 2
  21. 2025AnywhereDoor: Multi-Target Backdoor Attacks on Object Detectioncited by 2
  22. 2025Revisiting Training-Inference Trigger Intensity in Backdoor Attackscited by 2
  23. 2021CatchBackdoor: Backdoor Detection via Critical Trojan Neural Path Fuzzingcited by 1
  24. 2023Mask and Restore: Blind Backdoor Defense at Test Time with Masked Autoencodercited by 1
  25. 2023Pick your Poison: Undetectability versus Robustness in Data Poisoning Attackscited by 1
  26. 2024Evolutionary Trigger Detection and Lightweight Model Repair Based Backdoor Defensecited by 1
  27. 2026STEP: Detecting Audio Backdoor Attacks via Stability-based Trigger Exposure Profilingcited by 1
  28. 2022AntidoteRT: Run-time Detection and Correction of Poison Attacks on Neural Networkscited by 0
  29. 2022Contributor-Aware Defenses Against Adversarial Backdoor Attackscited by 0
  30. 2022Confidence Matters: Inspecting Backdoors in Deep Neural Networks via Distribution Transfercited by 0
  31. 2023Recover Triggered States: Protect Model Against Backdoor Attack in Reinforcement Learningcited by 0
  32. 2023Mitigating Backdoors within Deep Neural Networks in Data-limited Configurationcited by 0
  33. 2024BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learningcited by 0
  34. 2024A general approach to enhance the survivability of backdoor attacks by decision path couplingcited by 0
  35. 2024Hypnopaedia-Aware Machine Unlearning via Psychometrics of Artificial Mental Imagerycited by 0
  36. 2024AnywhereDoor: Multi-Target Backdoor Attacks on Object Detectioncited by 0
  37. 2024An Effective and Resilient Backdoor Attack Framework against Deep Neural Networks and Vision Transformerscited by 0
  38. 2025NT-ML: Backdoor Defense via Non-target Label Training and Mutual Learningcited by 0

50 most cited of 52 All 52 (JSON)