Classification of Malware Using VGGNet-19 Model – An Evaluation for General Understanding
Keywords:
Malware Classification, Deep Learning, Machine Learning, Artificial IntelligenceAbstract
The malware program becomes a disturbance to the user as it grabs the most important files and data of the user. The number of devices attached to the internet is increasing at a high speed which raises the opportunity for the attackers to steal or destroy the user data. Some malicious attackers demand money from the user after hacking the user's most important information. The traditional methods for malware classification including static and dynamic contain some limitations as they consume time in feature extraction. The malware classification needs to be identified in the light of artificial intelligence. The machine learning model is also used by the researcher for classification but now gets old and underperforms in the large dataset to overcome the large dataset issue deep learning algorithms are required which perform efficiently in large datasets. Difference researchers proposed deep learning algorithms for malware detection and achieving the best performance in detection but limited to malware classes. In this paper, the VGGNet-19 model which is also a deep learning model proposed for malware classification is in multiclass. In the VGGNet-19 model, the term VGG stands for Visual Geometric Group which can handle up to 19 layers for its deep network feature. The Malimg dataset contains 9339 samples divided into 25 malware classes. The proposed model is trained on this multiclass malware image dataset, achieving 0.998% performance in training, 0.990% in testing, and 0.990% in overall classification accuracy. The confusion matrix confirms model excellence performance across all classes.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2024 Humza Rana

This work is licensed under a Creative Commons Attribution 4.0 International License.
Copyright and Licensing
Publication is open access
Creative Commons Attribution License - CC BY- 4.0
Copyrights: The author retains unrestricted copyrights and publishing rights