Malware Classification Using Deep Learning Approaches

Detecting malware poses significant challenges due to the evolving sophistication of malicious software. Conventional detection techniques like signature-based methods and behavioral analysis have experienced diminishing effectiveness. To address this, a groundbreaking deep learning framework called Decomposition Training and Classification Network (DTC-Net) is introduced in this work. DTC-Net consists of three phases: feature extraction utilizing the VGG16 backbone network, data simplification through class decomposition, and classification using fully connected layers. Furthermore, a class composition layer enhances the accuracy of final categorization. Through extensive experimentation, the proposed DTC-Net demonstrates impressive performance, surpassing other state-of-the-art methods. It achieves remarkable results, including an F1-score of 95.364%, accuracy of 98.545%, precision of 94.737%, and recall of 96%. This research showcases the effectiveness of DTC-Net as an integrated and comprehensive deep learning solution for accurate malware family classification.