mlsanityData to 14 Oct 2026

Poison is Not Traceless: Fully-Agnostic Detection of Poisoning Attacks

arXiv 2310.16224 ↗First posted 24 Oct 2023cs.CR cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

The performance of machine learning models depends on the quality of the underlying data. Malicious actors can attack the model by poisoning the training data. Current detectors are tied to either specific data types, models, or attacks, and therefore have limited applicability in real-world scenarios. This paper presents a novel fully-agnostic framework, DIVA (Detecting InVisible Attacks), that detects attacks solely relying on analyzing the potentially poisoned data set. DIVA is based on the idea that poisoning attacks can be detected by comparing the classifier's accuracy on poisoned and clean data and pre-trains a meta-learner using Complexity Measures to estimate the otherwise unknown accuracy on a hypothetical clean dataset. The framework applies to generic poisoning attacks. For evaluation purposes, in this paper, we test DIVA on label-flipping attacks.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 29 of its 61 references that are in the map. First posted on arXiv 24 Oct 2023.

Builds on

29 of 61 references.

References · 29

  1. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,106
  2. 2018Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labelscited by 1,059
  3. 2018Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networkscited by 580
  4. 2018Detecting Backdoor Attacks on Deep Neural Networks by Activation Clusteringcited by 456
  5. 2016Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samplescited by 446
  6. 2017Wild Patterns: Ten Years After the Rise of Adversarial Machine Learningcited by 414
  7. 2020FLTrust: Byzantine-robust Federated Learning via Trust Bootstrappingcited by 346
  8. 2017Certified Defenses for Data Poisoning Attackscited by 303
  9. 2021Anti-Backdoor Learning: Training Clean Models on Poisoned Datacited by 263
  10. 2019Detecting AI Trojans Using Meta Neural Analysiscited by 199
  11. 2017Towards Poisoning of Deep Learning Algorithms with Back-gradient Optimizationcited by 197
  12. 2018Certified Adversarial Robustness with Additive Noisecited by 180
17 more
  1. 2019Universal Litmus Patterns: Revealing Backdoor Attacks in CNNscited by 146
  2. 2018Stronger Data Poisoning Attacks Break Data Sanitization Defensescited by 135
  3. 2020Deep Partition Aggregation: Provable Defense against General Poisoning Attackscited by 115
  4. 2020Certified Robustness to Label-Flipping Attacks via Randomized Smoothingcited by 108
  5. 2020Practical Detection of Trojan Neural Networks: Data-Limited and Data-Free Casescited by 95
  6. 2020Intrinsic Certified Robustness of Bagging against Data Poisoning Attackscited by 94
  7. 2022Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoningcited by 94
  8. 2018Is feature selection secure against training data poisoning?cited by 93
  9. 2019Bypassing Backdoor Detection Algorithms in Deep Learningcited by 74
  10. 2018Label Sanitization against Label Flipping Poisoning Attackscited by 73
  11. 2018How Complex is your classification problem? A survey on measuring classification complexitycited by 66
  12. 2018Detection of Adversarial Training Examples in Poisoning Attacks through Anomaly Detectioncited by 46
  13. 2021De-Pois: An Attack-Agnostic Defense against Data Poisoning Attackscited by 33
  14. 2019Spectral Metric for Dataset Complexity Assessmentcited by 15
  15. 2021Adversarially Robust Kernel Smoothingcited by 4
  16. 2023Fast Adversarial Label-Flipping Attack on Tabular Datacited by 3
  17. 2021BAARD: Blocking Adversarial Examples by Testing for Applicability, Reliability and Decidabilitycited by 2

Built on by

Not cited yet.