An explainable Polymorphic Malware Detection Framework with Transfer Learning in Encrypted Network
VFAST Transactions on Software Engineering · 20.06.2026 · 14 (2) · 263-280
DOI: 10.21015/vtse.v14i2.2406 ↗Kurz gefasst
Fachartikel in der Zeitschrift VFAST Transactions on Software Engineering, online erschienen am 20.06.2026. Verfasst von Ogba Paul Onu und 2 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben.
Abstract (Originalsprache)
This study focuses on the security of encrypted network traffic where problem of detecting polymorphic malware is paramount especially when traditional signature-based and machine learning methods fail due to detect code obfuscation and encryption. To overcome this, we propose a novel, comprehensive deep learning framework that employs hybrid feature extraction technique by converting raw traffic (PCAP) into both 2D grayscale images and temporal sequences. The framework is designed to be robust thereby using a Generative Adversarial Network (GAN) during model training and augmenting minority-class malware data to combat imbalance. It also hybridizes pre-trained VGG16 and GoogleNet CNNs extract spatial features from images, while an LSTM subnetwork analyzes temporal behavioral patterns. Detection and prediction are fused for final classification. Enhanced with a GRAD-CAM heatmap for explainability and operator trust, the model provides visual justification for its detections. We evaluate the model on the CICMalDroid 2020 benchmark dataset. The proposed model’s evaluation performance achieves an accuracy of 95%, 98.42% recall, 92.12% precision, a 95.17% F1-score, and a false positive rate of 8.42%. Comparative analysis of the results against recent works revealed that it outperformed several related frameworks. This means the proposed hybrid approach is effective in handling emerging malware challenges.
Zitieren
APA
Onu, O. P., Timothy, M., & Eke, C. I. (2026). An explainable Polymorphic Malware Detection Framework with Transfer Learning in Encrypted Network. VFAST Transactions on Software Engineering, 14(2), 263-280. https://doi.org/10.21015/vtse.v14i2.2406
Vancouver
Onu OP, Timothy M, Eke CI. An explainable Polymorphic Malware Detection Framework with Transfer Learning in Encrypted Network. VFAST Transactions on Software Engineering. 2026;14(2):263-280. doi:10.21015/vtse.v14i2.2406
BibTeX
@article{Onu2026,
author = {Onu, Ogba Paul and Timothy, Moses and Eke, Christopher Ifeanyi},
title = {An explainable Polymorphic Malware Detection Framework with Transfer Learning in Encrypted Network},
journal = {VFAST Transactions on Software Engineering},
year = {2026},
volume = {14},
number = {2},
pages = {263--280},
doi = {10.21015/vtse.v14i2.2406}
}
EUVIMED