Enhanced Image-Based Malware Multiclass Classification Method with the Ensemble Model and SVM

Abstract

Abstract Malware has become one of the biggest risks to security due to its rapid expansion. Therefore, it must be quickly detected and removed. While convolutional neural network (CNN) models have expanded to include ensemble and transfer learning approach from previous individual CNN architectures, relatively few studies have compared how well these approaches perform when it comes to malware family detection. A small number of malware varieties have been the focus of several research efforts’ studies. In this study, both of these issues were resolved. We present our own ensemble model for the classification of malware diseases into 34 types. We merge the Microsoft malware dataset with the Malimg dataset to increase the number of malware families identified by the model. To reduce training time and resource consumption, the suggested model utilized the most significant malware features, which are chosen based on the Least Absolute Shrinkage and Selection Operator method, for the purpose of classifying the malware classes. The experimental findings demonstrate that the ensemble model’s accuracy is 99.78%. Based on the experimental results, we conclude that the model will help with real-world malware classification tasks.