Malware Image Classification Using Global Context Vision Transformers for Information Security
ICCK Transactions on Information Security and Cryptography · 2025
DOI: 10.62762/tisc.2025.775760 ↗Kurz gefasst
Fachartikel in der Zeitschrift ICCK Transactions on Information Security and Cryptography, 2025. Verfasst von Muhammad Masab und 3 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref).
Abstract (Originalsprache)
The continuous threat of malware against digital systems exists because its attack methods develop rapidly, reducing the effectiveness of traditional detection systems. Current static and dynamic analysis methods for malware detection face challenges with scalability and robustness when handling large and complex malware samples. Computer vision now shows that malware binaries contain specific structural patterns when displayed as grayscale images, which can be used for classification. This study investigates GCViT for malware detection through its application to the Malimg dataset, which contains 9,337 samples from 25 malware families. The dataset underwent preprocessing through a two-step process that involved converting binary files into grayscale images followed by applying viridis colormap transformation and normalization for better visual discrimination. The GCViT model trained using ImageNet-pretrained weights while keeping its backbone fixed and modifying only the classifier head for malware family classification. The model reached 99.46% test accuracy and showed high effectiveness across most malware families, with only a few errors among structurally similar variants. The results demonstrate that GCViT achieves better performance by detecting both local and global dependencies in images, leading to improved malware image classification. The research sets a new benchmark for the Malimg dataset and highlights the potential of Vision Transformers in cybersecurity.
Zitieren
APA
Masab, M., Ahmad, K., Hussain, M., & Khan, M. S. (2025). Malware Image Classification Using Global Context Vision Transformers for Information Security. ICCK Transactions on Information Security and Cryptography. https://doi.org/10.62762/tisc.2025.775760
Vancouver
Masab M, Ahmad K, Hussain M, Khan MS. Malware Image Classification Using Global Context Vision Transformers for Information Security. ICCK Transactions on Information Security and Cryptography. 2025. doi:10.62762/tisc.2025.775760
BibTeX
@article{Masab2025,
author = {Masab, Muhammad and Ahmad, Khubab and Hussain, Muzammil and Khan, Muhammad Shahbaz},
title = {Malware Image Classification Using Global Context Vision Transformers for Information Security},
journal = {ICCK Transactions on Information Security and Cryptography},
year = {2025},
doi = {10.62762/tisc.2025.775760}
}
EUVIMED