Date of Award

7-7-2020

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Computer Science

First Advisor

Saeed Samet

Keywords

Deep Learning, Differential Privacy, Privacy-preserving

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

With recent technological advancements, the amount of personal user data that is being generated is immense. Due to the large volume of data, machine learning algorithms such as neural networks are serving as the backbone to derive patterns from this data quickly. This need for big data analytics comes at the cost of the privacy of user data. The second challenge that must be solved relates to the scalability of the machine learning algorithm. Neural networks are known to deteriorate as the volume of the data increases due to complex sum and sigmoid calculations. Therefore in this thesis, an attempt to parallelize the neural network while also maintaining the privacy of user data is made. This model would provide a viable option for big data analytics without sacrificing the privacy of individual users while also maintaining precision and the classification accuracy of the model. The implementation of the parallelized privacy preserving neural network will be based on the MapReduce computing model which provides advanced features such as fault tolerance, data replication, and load balancing.

Share

COinS