CNN-ViT synergy: An efficient Android malware detection approach through deep learning

The surge in malicious Android applications poses a significant risk to global smartphone security which demands robust detection strategies that are both effective and efficient. Traditional malware detection methods often rely on complex feature sets that can slow down analysis and obscure key insights. To simplify malware detection, this study presents a novel approach by converting network traffic data into images, which are then analyzed using deep learning models. We introduce hybrid models that seamlessly integrate Convolutional Neural Networks (CNN) and Vision Transformers (ViT) to capitalize on their respective strengths in identifying malicious traffic. Notably, our method explores various image resolutions, finding that a 180x180 resolution optimizes detection accuracy without compromising much processing speed. The proposed model achieves a groundbreaking 99.61% multiclass accuracy rate which demonstrates the effectiveness in distinguishing between benign and malicious applications with high precision. This research not only sets a new standard in Android malware detection efficiency but also paves the way for future advancements in the application of deep learning for cybersecurity.