An Efficient Malware Classification Model Using Convnets and GCN Techniques

Malware detection and classification remain critical tasks in cybersecurity due to the evolving nature of malicious software. In this paper, an efficient malware classification model is proposed, combining Convents and Graph Convolutional Networks (GCN) techniques. Leveraging the Kaggle dataset, which comprises 25 classes of Windows malware images, the aim is to develop a robust classification system. Initially, CNN is employed to extract features from the images, achieving an impressive accuracy of 96%. Recognizing the interconnectedness between malware samples, GCN, a powerful tool for modeling graph-structured data, is integrated to capture the relationships between different malware instances effectively, resulting in an accuracy of 98%. Subsequently, a novel integration of CNN and GCN is applied, boosting classification accuracy to 98.7%. This approach not only demonstrates good performance in malware classification but also showcases the effectiveness of combining CNN and GCN techniques. The proposed model provides a promising solution for malware detection and classification, thereby enhancing cybersecurity efforts in mitigating threats posed by malicious software.