mlsanityData to 14 Oct 2026

TrajSyn: Privacy-Preserving Dataset Distillation from Federated Model Trajectories for Server-Side Adversarial Training

arXiv 2512.15123 ↗First posted 17 Dec 2025cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Deep learning models deployed on edge devices are increasingly used in safety-critical applications. However, their vulnerability to adversarial perturbations poses significant risks, especially in Federated Learning (FL) settings where identical models are distributed across thousands of clients. While adversarial training is a strong defense, it is difficult to apply in FL due to strict client-data privacy constraints and the limited compute available on edge devices. In this work, we introduce TrajSyn, a privacy-preserving framework that enables effective server-side adversarial training by synthesizing a proxy dataset from the trajectories of client model updates, without accessing raw client data. We show that TrajSyn consistently improves adversarial robustness on image classification benchmarks with no extra compute burden on the client device.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 35 references that are in the map. First posted on arXiv 17 Dec 2025.

Builds on

27 of 35 references.

References · 27

  1. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  3. 2013Intriguing properties of neural networkscited by 5,155
  4. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 3,104
  5. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,340
  6. 2015Distillation as a Defense to Adversarial Perturbations against Deep Neural Networkscited by 678
  7. 2017Provable defenses against adversarial examples via the convex outer adversarial polytopecited by 567
  8. 2017Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methodscited by 546
  9. 2020Dataset Condensation with Gradient Matchingcited by 516
  10. 2018Efficient Neural Network Robustness Certification with General Activation Functionscited by 445
  11. 2022Dataset Distillation by Matching Training Trajectoriescited by 442
  12. 2018Dynamic Few-Shot Visual Learning without Forgettingcited by 385
15 more
  1. 2021Dataset Condensation with Differentiable Siamese Augmentationcited by 291
  2. 2018Sorting out Lipschitz function approximationcited by 213
  3. 2017Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulationcited by 185
  4. 2018Certified Adversarial Robustness with Additive Noisecited by 180
  5. 2023GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Modelscited by 176
  6. 2020Distilled One-Shot Federated Learningcited by 109
  7. 2019Soft-Label Dataset Distillation and Text Dataset Distillationcited by 107
  8. 2023Black Box Adversarial Prompting for Foundation Modelscited by 62
  9. 2018Provable Robustness of ReLU networks via Maximization of Linear Regionscited by 51
  10. 2022Federated Learning via Decentralized Dataset Distillation in Resource-Constrained Edge Environmentscited by 49
  11. 2020FAT: Federated Adversarial Trainingcited by 24
  12. 2023Adversarial Attacks on Traffic Sign Recognition: A Surveycited by 13
  13. 2024Recent Advances in Attack and Defense Approaches of Large Language Modelscited by 6
  14. 2025XOXO: Stealthy Cross-Origin Context Poisoning Attacks against AI Coding Assistantscited by 4
  15. 2025AdvSumm: Adversarial Training for Bias Mitigation in Text Summarizationcited by 2

Built on by

Not cited yet.