Electronic Theses and Dissertations

Date of Award

8-1-2022

Document Type

Thesis

Degree Name

M.S. in Engineering Science

First Advisor

Charles Fleming

Second Advisor

Yixin Chen

Third Advisor

Charlie Walter

School

University of Mississippi

Relational Format

dissertation/thesis

Abstract

In recent years, the amount of data being produced has increased tremendously. This data has allowed us to create and train machine learning models that are being used nowadays. Though these models have proven to be very effective in classification and regression problems, they do have a vulnerability. This vulnerability is exploited by adding adversarial perturbations to the original data or in the deployed model. These adversarial attacks are done by making minute changes to the data to confuse the machine learning model. This can lead to models’ misrepresentation and drop the accuracy. For the span of this paper, I have used various statistical and visualization analysis methods to find the effect of adversarial perturbations on the CIFAR10 image data. Some of these methods involved were mean, standard deviation, variance, covariance, the probability distribution of color channels, etc. This paper discusses the insights found during the analysis of the CIFAR10 image dataset and the future work to be expected in this field.

Concentration/Emphasis

Computer Sciences

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.