Proceedings of the 8th World Conference on Future of Education
Year: 2024
DOI:
[PDF]
Image Based Improved Convolutional Neural Network Towards Malware Classification
Gregory .O. Onwodi, Kennedy .E. Ketebu
ABSTRACT:
Computer Security is a critical consideration in the era of widespread computer usage. Among the myriad of threats are malwares. Malware are simply malicious programs or files that compromise the security of computer systems, leading to information loss, data theft, service disruption, and financial losses. The evolving nature of malware makes detection and classification challenging, as evidenced by the continuous rise in malware attacks. This paper introduces an innovative approach to malware classification using an improved convolutional neural network (I-CNN). Leveraging the effectiveness of convolutional neural networks (CNNs) which eliminates the need to disassemble or execute malware binaries during malware analysis but uses visualization to identify local and global descriptor features of malware images. The development and evaluation of I-CNN were performed using the hyperparameter tuning on Malevis dataset. Additionally, the model’s cross dataset generalization ability was tested on the Malimg dataset, comparing its performance against other existing researcher’s models. The result showed I-CNN, achieving an accuracy score of 98.88%. This highlights the potential of image-based classification models, especially I-CNN, in addressing the evolving challenges posed by malwares in computer systems. This work helps to protect Computer Systems deployed in Learning Technologies thereby securing the future of education which is largely anchored on Learning Technologies.
keywords: Malevis, Dataset, Computer-security, hyperparameter, Malimg