mlsanityData to 7 Oct 2026

Reconstructing Training Data from Multiclass Neural Networks

arXiv 2305.03350 ↗First posted 5 May 2023cs.LG cs.CR cs.CVcited by 7

Not on the map. Why?

Abstract (arXiv)

Reconstructing samples from the training set of trained neural networks is a major privacy concern. Haim et al. (2022) recently showed that it is possible to reconstruct training samples from neural network binary classifiers, based on theoretical results about the implicit bias of gradient methods. In this work, we present several improvements and new insights over this previous work. As our main improvement, we show that training-data reconstruction is possible in the multi-class setting and that the reconstruction quality is even higher than in the case of binary classification. Moreover, we show that using weight-decay during training increases the vulnerability to sample reconstruction. Finally, while in the previous work the training set was of size at most 1000 from 10 classes, we show preliminary evidence of the ability to reconstruct from a model trained on 5000 samples from 100 classes.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Builds on 13 of its 21 references that are in the map. First posted on arXiv 5 May 2023.

Builds on

13 of 21 references.

References · 13

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  2. 2015Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classificationcited by 3,590
  3. 2020Extracting Training Data from Large Language Modelscited by 2,144
  4. 2018The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networkscited by 847
  5. 2023Extracting Training Data from Diffusion Modelscited by 723
  6. 2017The Implicit Bias of Gradient Descent on Separable Datacited by 649
  7. 2018Exploiting Unintended Feature Leakage in Collaborative Learningcited by 509
  8. 2017Deep Models Under the GAN: Information Leakage from Collaborative Deep Learningcited by 326
  9. 2019Gradient Descent Maximizes the Margin of Homogeneous Neural Networkscited by 308
  10. 2021Evaluating Gradient Inversion Attacks and Defenses in Federated Learningcited by 196
  11. 2020Directional convergence and alignment in deep learningcited by 172
  12. 2022Reconstructing Training Data with Informed Adversariescited by 143
1 more
  1. 2022Reconstructing Training Data from Trained Neural Networkscited by 129

Built on by

Cited by year

Citing papers per year: 2023: 1; 2024: 4; 2025: 2.

7 papers.

  1. 2024Provable Privacy Attacks on Trained Shallow Neural Networkscited by 4
  2. 2025Shortcut Learning Susceptibility in Vision Classifierscited by 4
  3. 2023Fantastic DNN Classifiers and How to Identify them without Datacited by 1
  4. 2024Network Inversion for Training-Like Data Reconstructioncited by 1
  5. 2025Securing AI Systems: A Guide to Known Attacks and Impactscited by 1
  6. 2024Novel Deep Neural Network Classifier Characterization Metrics with Applications to Dataless Evaluationcited by 0
  7. 2024Network Inversion and Its Applicationscited by 0