Enhanced Android Malware Detect Models Based on Explainable Generative Adversarial Networks

Enhanced Android Malware Detect Models Based on Explainable Generative Adversarial Networks

With the widespread adoption of smartphones over the past decade, mobile applications have become a primary target for malicious attacks, usually in the form of malware. Recent studies have leveraged artificial intelligence (AI) techniques for malware detection and classification. However, applying...

Full description

Saved in:
Bibliographic Details
Main Authors: Shamma Alshebli, Hyeran Mun, Deepak Puthal, Mohamed Jamal Zemerly, Luigi Martino, Ernesto Damiani, Chan Yeob Yeun
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Malware classification
explainable artificial intelligence (XAI)
generative adversarial networks (GAN)
data augmentation
machine learning (ML)
deep learning (DL)
Online Access:https://ieeexplore.ieee.org/document/11062843/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the widespread adoption of smartphones over the past decade, mobile applications have become a primary target for malicious attacks, usually in the form of malware. Recent studies have leveraged artificial intelligence (AI) techniques for malware detection and classification. However, applying such approaches, particularly deep learning (DL) techniques, to mobile malware detection poses significant challenges. These challenges arise from the difficulty of collecting large quantities of mobile malware samples and the inherent class imbalance in the collected datasets. To tackle these issues and enhance the performance of machine learning (ML) and DL detection models, we propose novel detection models based on a generative adversarial network (GAN). Furthermore, our approach not only employs a conditional tabular GAN (CTGAN) for data augmentation to explore the impact of augmentation but also identifies the optimal multiplication factor for achieving the best results. The evaluation results demonstrate that the proposed data augmentation approach significantly improves the performance of mobile malware detection models, especially those based on DL. We have notably figured out that doubling the original dataset is sufficient to enhance the performance of ML models, whereas DL models require additional data to achieve optimal results. Hence, our proposed mechanism is an effective solution for improving mobile malware detection.
ISSN:2169-3536

Similar Items