EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

ARSNet: A Novel Malware Visualization Detection Method

Xinbo Liu, Wei Liu, Peng Nie, Zhuojun Fu

IEEE Access · 2026 · 14 · 85517-85530

DOI: 10.1109/access.2026.3696941 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Xinbo Liu und 3 weiteren Autor:innen. Die Arbeit nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Hunan Province Key Education Reform Project, Hunan Province Academic Degree and Graduate Education Reform Project, University-Level Project of Guangdong University of Science and Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Malware detection remains a significant challenge in cybersecurity. To address this, visualization-based methods map binary code into image representations to leverage visual features for rapid identification. This method has emerged as a highly promising paradigm for automated threat detection in cyberspace security. Modern deep learning models generally achieve high overall accuracy. However, this metric often masks severe performance degradation in long-tailed distribution scenarios. In such contexts, even state-of-the-art Vision Transformers (ViTs) fail to generalize to data-scarce minority variants. To address this critical class imbalance, this work proposes a novel method named ARSNet for rapid malware detection. By integrating the BASA module with the DHM loss function, ARSNet effectively extracts discriminative fine-grained textures and adaptively re-balances learning weights for hard samples. Experimental results demonstrate that ARSNet resolves the critical bottleneck of data scarcity while maintaining a highly competitive overall accuracy of 98.6%. Furthermore, compared to advanced baselines and state-of-the-art methods, ARSNet improves the recall of minority families by over 23%. This improvement provides a robust defense against imbalanced malware threats.

Förderung

  • Hunan Province Key Education Reform Project
  • Hunan Province Academic Degree and Graduate Education Reform Project
  • University-Level Project of Guangdong University of Science and Technology

Zitieren

APA

Liu, X., Liu, W., Nie, P., & Fu, Z. (2026). ARSNet: A Novel Malware Visualization Detection Method. IEEE Access, 14, 85517-85530. https://doi.org/10.1109/access.2026.3696941

Vancouver

Liu X, Liu W, Nie P, Fu Z. ARSNet: A Novel Malware Visualization Detection Method. IEEE Access. 2026;14:85517-85530. doi:10.1109/access.2026.3696941

BibTeX
@article{Liu2026,
  author = {Liu, Xinbo and Liu, Wei and Nie, Peng and Fu, Zhuojun},
  title = {ARSNet: A Novel Malware Visualization Detection Method},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {85517--85530},
  doi = {10.1109/access.2026.3696941}
}
Kostenlos starten →