Improving privacy of Federated Learning Generative Adversarial Networks using Intel SGX

Improving privacy of Federated Learning Generative Adversarial Networks using Intel SGX

Jehee, Wouter (TU Delft Electrical Engineering, Mathematics and Computer Science)

Liang, K. (mentor)
Urbano, Julián (graduation committee)
Wang, R. (mentor)

Delft University of Technology

Computer Science and Engineering

CSE3000 Research Project

2022-06-22

Federated learning (FL), although a major privacy improvement over centralized learning, is still vulnerable to privacy leaks. The research presented in this paper provides an analysis of the threats to FL Generative Adversarial Networks. Furthermore, an implementation is provided to better protect the data of the participants with Trusted Execution Environments (TEEs), using Intel Software Guard Extensions. Lastly, the viability of it’s use in practice is evaluated and discussed. The results indicate that this approach protects the data, while not affecting the predicting capabilities of the model, with a noticeable but manageable impact on the training duration.

Federated learning
Privacy-preserving
Intel SGX
Trusted Execution Environments
Generative Adversarial Networks

http://resolver.tudelft.nl/uuid:d30df7e9-a945-4c29-adb3-2273430a58a2

Student theses

bachelor thesis

© 2022 Wouter Jehee