Graduate Thesis Or Dissertation
 

Resource-Aware Distributed Data Sanitization for Privacy Preserving Machine Learning

Público Deposited

Contenido Descargable

Descargar PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/8c97kx80c

Descriptions

Attribute NameValues
Creator
Abstract
  • When a single object is captured by multiple edge devices, the data captured by every edge device could contain both public and private information. The private information in every copy of the data captured has to be preserved while the public information has to be utilized for any further machine learning tasks. We present a machine learning framework that achieves these two competing objectives in a distributed and resource constrained environment. In this work, we develop a data sanitizing algorithm as an interplay between the two competing objectives and address the problem of resource constraints by utilizing an iterative model pruning technique. To demonstrate the system’s performance, we conduct experiments on a custom MNIST dataset. We synthesize a two digit dataset of numbers in the range 00-19. Any even number in this dataset is considered as public information and any number in the range 10-19 as private information.
License
Resource Type
Fecha de Emisión
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Declaración de derechos
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2021-11-16 to 2023-12-16

Relaciones

Parents:

This work has no parents.

En Collection:

Elementos

Miniatura Título Fecha de subida Visibilidad Acciones
file details AnandYashwanthi2021.pdf 2021-11-15 Público Descargar