Cross-Platform Malware Classification: Fusion of CNN and GRU Models
Nagababu Pachhala, Subbaiyan Jothilakshmi, Bhanu Prakash Battula · International Journal of Safety and Security Engineering · 2024
Effective cross-platform malware categorization techniques are becoming more and more necessary as malware spreads across more systems.Conventional methods are primarily concerned with the static or dynamic aspects of malware, which often restricts their ability to identify and categorize malware on various operating systems.In this paper, we use both static and dynamic characteristics to present a unique deep learning-based method for cross-platform malware classification.Our work aims to identify the distinct features of malware on different operating systems, such as Windows, macOS, Android, and iOS.We provide a complete depiction of malware behavior by collecting both dynamic and static data, such as system calls and network traffic patterns, as well as file properties, API calls, and header information.Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) are two components of our deep learning architecture that we use to address the inherent issues of cross-platform malware categorization.This fusion of networks enables us to effectively capture both spatial and temporal patterns present in malware samples, enhancing the accuracy of classification across platforms.To evaluate the performance of our proposed model, we employ benchmark datasets encompassing diverse malware families across different operating systems.The results demonstrate superior classification accuracy, precision, recall, and F-score compared to traditional machine learning approaches and single-feature-based models.