This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/115643
Description
Title
Secure and scalable robust federated learning
Author(s)
Liu, Andrew
Issue Date
2022-04-26
Director of Research (if dissertation) or Advisor (if thesis)
Khurana, Dakshita
Department of Study
Computer Science
Discipline
Computer Science
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
federated learning
secure multi-party computation
privacy preserving machine learning
Abstract
With the rise of cloud computing in recent years, an increasing amount of data storage and computation are being offloaded from individual users’ computers to those of large corporations. These corporations benefit from merging their clients’ data, e.g. hospitals sharing client chest x-ray images to better detect and treat COVID-19. However, these institutions are restricted from sharing their client data due to both ethical and legal reasons. Federated learning is a technique that attempts to solve this problem by training a model across multiple edge devices while data stays on-device. This thesis focuses on the specific case of robust federated learning, where the central server coordinating the parties computes a robust aggregate of client updates to redistribute back as a global model. Cryptographic protocols such as secure multi-party computation and differential privacy are often added on top of federated learning to formally prove security. This work focuses on developing an efficient and private version of federated learning, when coordinate-wise median is used as a robust aggregator. A semi-honest protocol is developed for both median computation and approximate median computation. The convergence and robustness of this protocols is evaluated empirically.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.