Abstract

The detection of malware is an important challenge in the field of cybersecurity, which is demanding efficient and accurate solutions for making threats easier. This study is focused on the detection of malware by integrating a framework that uses a Mamdani fuzzy interference system (FIS) and diffusion techniques based on three different deep learning algorithms, which are convolutional neural networks (CNNs), recurrent neural networks (RNNs), and generative adversarial networks (GANs) for malware detection by using recent dataset. This was done by using their outputs through a voting and routing mechanism. The results of the study show that the fuzzy logic system was able to get the highest accuracy of 97.68% and F1-score of 97.31%, which shows that it performed better than the individual models. The CNN model also showed a strong performance which had an accuracy of 93.08%, while the RNN and GAN models achieved accuracies of 90.14% and 89.43%. Evaluation metrics such as precision, recall, and ROC-AUC highlighted the fuzzy logic system's robustness for the detection of malware while also minimizing the false positives and the negatives. These findings show the potential of the hybrid approaches which is combining machine learning models and fuzzy logic to enhance the malware detection frameworks.

Library of Congress Subject Headings

Malware (Computer software)--Identification--Data processing; Malware (Computer software)--Prevention--Data processing; Deep learning (Machine learning); Neural networks (Computer science)

Publication Date

12-19-2024

Document Type

Thesis

Student Type

Graduate

Degree Name

Computing Security (MS)

Advisor

Kevser Ovaz Akpinar

Advisor/Committee Member

Huda Saadeh

Advisor/Committee Member

Ali Assi

Campus

RIT Dubai

Plan Codes

COMPSEC-MS

Download

Share

COinS