Skip to content

Project about privacy in Machine Learning for the course 02460 Advanced Machine Learning Spring 2021

Notifications You must be signed in to change notification settings

madsemildahlgaard/dlg-attack-configuration

 
 

Repository files navigation

Influence of attack configurations on the reconstruction of medical images in federated learning

Code for paper under review for MIDL 2022 about privacy in federated machine learning.

We explore the robustness and techniques of deep leakage from gradients (DLG) [1]. We investigate the influence of different initalization methods and distance measures by comparing the convergence rate and speed of single image reconstructions. Specifically, we compare the results with SAPAG [2].

Algorithm

DLG illustration

References

[1] Zhu et. al (2020). "Deep Leakage from Gradients” Lecture Notes in Computer Science (including sub-series Lecture Notes in Artificial Intelligence and LectureNotes in Bioinformatics), vol. 12500 LNCS, no. NeurIPS,pp. 17–31.

[2] Wang et. al. (2020). “SAPAG: A self-adaptive privacy attackfrom gradients,”. arXiv.

About

Project about privacy in Machine Learning for the course 02460 Advanced Machine Learning Spring 2021

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 93.9%
  • Python 6.0%
  • Shell 0.1%