Date of Award

2023

Document Type

Honors

Degree Name

General Beadle Honors Program

Department

Computer Science

First Advisor

Andrew Kramer

Abstract

AI usage in our society is more prevalent today compared to years ago. Our evolution in computing has created problems that never existed before because of a fundamental lack of security-minded development. Security was always a second thought and never a priority. With the prevalence of AI making security a bigger priority, the question arises if AIs such as ChatGPT, Github Co-pilot, Alpha Go, and Bing AI use the best security practices when implemented. The creation of modern technologies such as phones, the Internet, and computers have all seen horrendous security concerns addressed later on in development. With how technology currently stands, there is always innovation, but security research consistently falls behind or is not considered. The world is always changing and finding new vulnerabilities constantly, especially with the speed of AI’s current development, which does not definitively address these security concerns. This project will identify different points of the development process where AI is susceptible to attacks or has vulnerabilities and create a framework to mitigate these concerns. The framework will also help developers enumerate these vulnerabilities and identify security fixes or considerations when developing AI. This proposed framework aims to help developers build and maintain more secure AI throughout the implementation process. This foundational knowledge will also help researchers in the field, as it establishes points to investigate for future projects.

Download

Share

COinS