Federated Learning Privacy

Federated Learning (FL) allows training to be performed in a distributed manner. This is however vulnerable to inference attacks or model poisoning. We study novel techniques to perform attacks, defenses and mitigations.

Publications

1. Gradient Inversion of Federated Diffusion Models
   • Conference: ARES 2025
   • Authors: Jiyue Huang, Chi Hong, Stefanie Roos and Lydia Y. Chen

   • ��Paper

     ��Code

2. TS-Inverse: A Gradient Inversion Attack Tailored for Federated Time Series Forecasting Models
   • Conference: SaTML 2025
   • Authors: Caspar Meijer, Jiyue Huang, Shreshtha Sharma, Elena Lazovik, Lydia Y Chen

   • ��Paper

     ��Code

3. On Quantifying the Gradient Inversion Risk of Data Reuse in Federated Learning Systems
   • Conference: ISRDS 2024
   • Authors: Jiyue Huang, Lydia Y. Chen, Stefanie Roos
   • ��Code
4. Fabricated Flips: Poisoning Federated Learning without Data
   • Conference: DSN 2023
   • Authors: Jiyue Huang, Zilong Zhao, Lydia Y. Chen, Stefanie Roos

   • ��Paper

     ��Code

5. Exploring and Exploiting Data-Free Model Stealing
   • Conference: DSN 2023
   • Authors: Chi Hong, Jiyue Huang, Robert Birke & Lydia Y. Chen
   • ��Paper
6. Defending Against Free-Riders Attacks in Distributed Generative Adversarial Networks
   • Conference: FC 2023
   • Authors: Zilong Zhao, Jiyue Huang, Lydia Y Chen, Stefanie Roos

   • ��Paper

     ��Code

7. LeadFL: Client Self-Defense against Model Poisoning in Federated Learning
   • Conference: ICML 2023
   • Authors: Chaoyi Zhu, Stefanie Roos, Lydia Y. Chen

   • ��Paper

     ��Code

8. AGIC: Approximate Gradient Inversion Attack on Federated Learning
   • Conference: SRDS 2022
   • Authors: Jin Xu, Chi Hong, Jiyue Huang, Lydia Y. Chen, Jérémie Decouchant
   • ��Paper