Date of Award

10-27-2021

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

School of Information Technology: Information Systems

First Advisor

Shaoen Wu

Abstract

Security vulnerabilities in source code are traditionally detected manually by software developers because there are no effective auto-detection tools. Current vulnerability detection tools require great human effort, and the results have flaws in many ways. However, deep learning models could be a solution to this problem for the following reasons: 1. Deep learning models are relatively accurate for text classification and text summarization for source code. 2. After being deployed on the cloud servers, the efficiency of deep learning based auto-detection could be much higher than human effort. Therefore, we developed two Natural Language Processing(NLP) models: the first one is a text-classification model that takes source code as input and outputs the classification of the security vulnerability of the input. The second one is a text-to-text model that takes source code as input and outputs a completely machine-generated summary about the security vulnerability of the input. Our evaluation shows that both models get impressive results.

Comments

Imported from Zhang_ilstu_0092N_12059.pdf

DOI

https://doi.org/10.30707/ETD2021.20220215070319220232.999972

Page Count

36

Download

Share

COinS