mlsanityData to 7 Oct 2026

Towards Eliminating Hard Label Constraints in Gradient Inversion Attacks

arXiv 2402.03124 ↗First posted 5 Feb 2024cs.CR cs.CV cs.LGcited by 4

Not on the map. Why?

Abstract (arXiv)

Gradient inversion attacks aim to reconstruct local training data from intermediate gradients exposed in the federated learning framework. Despite successful attacks, all previous methods, starting from reconstructing a single data point and then relaxing the single-image limit to batch level, are only tested under hard label constraints. Even for single-image reconstruction, we still lack an analysis-based algorithm to recover augmented soft labels. In this work, we change the focus from enlarging batchsize to investigating the hard label constraints, considering a more realistic circumstance where label smoothing and mixup techniques are used in the training process. In particular, we are the first to initiate a novel algorithm to simultaneously recover the ground-truth augmented label and the input feature of the last fully-connected layer from single-input gradients, and provide a necessary condition for any analytical-based label recovery methods. Extensive experiments testify to the label recovery accuracy, as well as the benefits to the following image reconstruction. We believe soft labels in classification tasks are worth further attention in gradient inversion attacks.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 30 of its 43 references that are in the map. First posted on arXiv 5 Feb 2024.

Builds on

30 of 43 references.

References · 30

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  5. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  6. 2022Generative Adversarial Networkscited by 5,603
  7. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  8. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  9. 2019Deep Leakage from Gradientscited by 1,004
  10. 2020Inverting Gradients -- How easy is it to break privacy in federated learning?cited by 707
  11. 2019Dreaming to Distill: Data-free Knowledge Transfer via DeepInversioncited by 419
  12. 2020iDLG: Improved Deep Leakage from Gradientscited by 367
18 more
  1. 2021See through Gradients: Image Batch Recovery via GradInversioncited by 320
  2. 2021No Fear of Heterogeneity: Classifier Calibration for Federated Learning with Non-IID Datacited by 241
  3. 2021Evaluating Gradient Inversion Attacks and Defenses in Federated Learningcited by 196
  4. 2020Privacy and Robustness in Federated Learning: Attacks and Defensescited by 163
  5. 2021Gradient Inversion with Generative Image Priorcited by 101
  6. 2020R-GAP: Recursive Gradient Attack on Privacycited by 95
  7. 2022Auditing Privacy Defenses in Federated Learning via Generative Gradient Leakagecited by 72
  8. 2020A Framework for Evaluating Gradient Leakage Attacks in Federated Learningcited by 63
  9. 2022GradViT: Gradient Inversion of Vision Transformerscited by 52
  10. 2022Towards Understanding and Mitigating Dimensional Collapse in Heterogeneous Federated Learningcited by 45
  11. 2021Towards General Deep Leakage in Federated Learningcited by 40
  12. 2022A Survey on Gradient Inversion: Attacks, Defenses and Future Directionscited by 39
  13. 2021Bayesian Framework for Gradient Leakagecited by 34
  14. 2020Rethinking Privacy Preserving Deep Learning: How to Evaluate and Thwart Privacy Attackscited by 32
  15. 2020SAPAG: A Self-Adaptive Privacy Attack From Gradientscited by 27
  16. 2021Revealing and Protecting Labels in Distributed Trainingcited by 17
  17. 2020Exploring the Security Boundary of Data Reconstruction via Neuron Exclusivity Analysiscited by 10
  18. 2024Exploring Vacant Classes in Label-Skewed Federated Learningcited by 10

Built on by

Cited by year

Citing papers per year: 2025: 4.

4 papers.

  1. 2025Exploring the Vulnerabilities of Federated Learning: A Deep Dive into Gradient Inversion Attackscited by 9
  2. 2025Effective and Efficient Cross-City Traffic Knowledge Transfer: A Privacy-Preserving Perspectivecited by 0
  3. 2025FedTDP: A Privacy-Preserving and Unified Framework for Trajectory Data Preparation via Federated Learningcited by 0
  4. 2025MAGIA: Sensing Per-Image Signals from Single-Round Averaged Gradients for Label-Inference-Free Gradient Inversioncited by 0