mlsanityData to 7 Oct 2026

Topological Detection of Trojaned Neural Networks

arXiv 2106.06469 ↗First posted 11 Jun 2021cs.LG cs.CGcited by 34

Not on the map. Why?

Abstract (arXiv)

Deep neural networks are known to have security issues. One particular threat is the Trojan attack. It occurs when the attackers stealthily manipulate the model's behavior through Trojaned training samples, which can later be exploited. Guided by basic neuroscientific principles we discover subtle -- yet critical -- structural deviation characterizing Trojaned models. In our analysis we use topological tools. They allow us to model high-order dependencies in the networks, robustly compare different networks, and localize structural abnormalities. One interesting observation is that Trojaned models develop short-cuts from input to output layers. Inspired by these observations, we devise a strategy for robust detection of Trojaned models. Compared to standard baselines it displays better performance on multiple benchmarks.

Key facts. Cited by 34 papers in the map (data through 7 Oct 2026). Builds on 43 of its 75 references that are in the map. First posted on arXiv 11 Jun 2021.

Builds on

43 of 75 references.

References · 43

  1. 2017Attention Is All You Needcited by 59,932
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  3. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  4. 2016Densely Connected Convolutional Networkscited by 6,691
  5. 2013Intriguing properties of neural networkscited by 5,114
  6. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,052
  7. 2017Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoningcited by 1,012
  8. 2019STRIP: A Defence Against Trojan Attacks on Deep Neural Networkscited by 437
  9. 2018Detecting Backdoor Attacks on Deep Neural Networks by Activation Clusteringcited by 429
  10. 2020Weight Poisoning Attacks on Pre-trained Modelscited by 318
  11. 2019Label-Consistent Backdoor Attackscited by 285
  12. 2015Persistence Images: A Stable Vector Representation of Persistent Homologycited by 235
31 more
  1. 2019Topology-Preserving Deep Image Segmentationcited by 198
  2. 2017Neural Trojanscited by 151
  3. 2019Universal Litmus Patterns: Revealing Backdoor Attacks in CNNscited by 143
  4. 2018Backdoor Embedding in Convolutional Neural Network Models via Invisible Perturbationcited by 131
  5. 2019PersLay: A Neural Network Layer for Persistence Diagrams and New Graph Topological Signaturescited by 125
  6. 2019TABOR: A Highly Accurate Approach to Inspecting and Restoring Trojan Backdoors in AI Systemscited by 120
  7. 2019Ripser: efficient computation of Vietoris-Rips persistence barcodescited by 115
  8. 2017Deep Learning with Topological Signaturescited by 113
  9. 2020Topology of deep neural networkscited by 111
  10. 2020Practical Detection of Trojan Neural Networks: Data-Limited and Data-Free Casescited by 93
  11. 2018Technical Report: When Does Machine Learning FAIL? Generalized Transferability for Evasion and Poisoning Attackscited by 78
  12. 2017Sliced Wasserstein Kernel for Persistence Diagramscited by 77
  13. 2020A Topological Filter for Learning with Label Noisecited by 76
  14. 2021Topology-Aware Segmentation Using Discrete Morse Theorycited by 75
  15. 2014A Stable Multi-Scale Kernel for Topological Machine Learningcited by 74
  16. 2018Neural Persistence: A Complexity Measure for Deep Neural Networks Using Algebraic Topologycited by 73
  17. 2019Graph Filtration Learningcited by 68
  18. 2018A Topological Regularizer for Classifiers via Persistent Homologycited by 62
  19. 2019Learning metrics for persistence-based summaries and applications for graph classificationcited by 57
  20. 2020Computing the Testing Error without a Testing Setcited by 54
  21. 2019Connectivity-Optimized Representation Learning via Persistent Homologycited by 43
  22. 2021Link Prediction with Persistent Homology: An Interactive Viewcited by 37
  23. 2019TrojDRL: Trojan Attacks on Deep Reinforcement Learning Agentscited by 29
  24. 2021Topological Uncertainty: Monitoring trained neural networks through persistence of activation graphscited by 20
  25. 2018Topological Data Analysis of Decision Boundaries with Application to Model Selectioncited by 19
  26. 2013Confidence sets for persistence diagramscited by 17
  27. 2021TopoTxR: A Topological Biomarker for Predicting Treatment Response in Breast Cancercited by 13
  28. 2017Adversary Detection in Neural Networks via Persistent Homologycited by 12
  29. 2020Finding the Homology of Decision Boundaries with Active Learningcited by 12
  30. 2020Towards Interaction Detection Using Topological Analysis on Neural Networkscited by 5
  31. 2019Computing Minimal Persistent Cycles: Polynomial and Hard Casescited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 3; 2022: 8; 2023: 10; 2024: 8; 2025: 3; 2026: 2.

34 papers.

Sort:

By area:

  1. 2022Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networkscited by 126
  2. 2024Position: Topological Deep Learning is the New Frontier for Relational Learningcited by 65
  3. 2022A Study of the Attention Abnormality in Trojaned BERTscited by 47
  4. 2023UNICORN: A Unified Backdoor Trigger Inversion Frameworkcited by 44
  5. 2021Trigger Hunting with a Topological Prior for Trojan Detectioncited by 39
  6. 2023NOTABLE: Transferable Backdoor Attacks Against Prompt-based NLP Modelscited by 33
  7. 2023Attention-Enhancing Backdoor Attacks Against BERT-based Modelscited by 32
  8. 2021Cycle Representation Learning for Inductive Relation Predictioncited by 23
  9. 2023Backdoor Attacks Against Dataset Distillationcited by 20
  10. 2023Defending Against Patch-based Backdoor Attacks on Self-Supervised Learningcited by 19
  11. 2022A Survey of Neural Trojan Attacks and Defenses in Deep Learningcited by 17
  12. 2022Neural Approximation of Graph Topological Featurescited by 16
22 more
  1. 2023BEAGLE: Forensics of Deep Learning Backdoor Attack for Better Defensecited by 15
  2. 2023Defenses in Adversarial Machine Learning: A Surveycited by 14
  3. 2024Preference Poisoning Attacks on Reward Model Learningcited by 14
  4. 2022Backdoor Attack and Defense in Federated Generative Adversarial Network-based Medical Image Synthesiscited by 12
  5. 2021Check Your Other Door! Creating Backdoor Attacks in the Frequency Domaincited by 9
  6. 2023Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNscited by 8
  7. 2023Look, Listen, and Attack: Backdoor Attacks Against Video Action Recognitioncited by 7
  8. 2022Attention Hijacking in Trojan Transformerscited by 6
  9. 2025DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Datacited by 5
  10. 2022On the Convergence of Optimizing Persistent-Homology-Based Lossescited by 2
  11. 2024Mixup Barcodes: Quantifying Geometric-Topological Interactions between Point Cloudscited by 2
  12. 2022Model-Contrastive Learning for Backdoor Defensecited by 1
  13. 2023Mask and Restore: Blind Backdoor Defense at Test Time with Masked Autoencodercited by 1
  14. 2024Be Persistent: Towards a Unified Solution for Mitigating Shortcuts in Deep Learningcited by 1
  15. 2024Trojan Cleansing with Neural Collapsecited by 1
  16. 2025Topology of Out-of-Distribution Examples in Deep Neural Networkscited by 1
  17. 2024CBPF: Filtering Poisoned Data Based on Composite Backdoor Attackcited by 0
  18. 2024Characterization of topological structures in different neural network architecturescited by 0
  19. 2024Hypnopaedia-Aware Machine Unlearning via Psychometrics of Artificial Mental Imagerycited by 0
  20. 2025ArcGen: Generalizing Neural Backdoor Detection Across Diverse Architecturescited by 0
  21. 2026Detecting Semantic Backdoors in a Mystery Shopping Scenariocited by 0
  22. 2026Delayed Backdoor Attacks: Exploring the Temporal Dimension as a New Attack Surface in Pre-Trained Modelscited by 0