Image Classification Pada Malware Menggunakan Pendekatan Metode Deep Learning VGG-16

Implementation of Malware is designed to damage hardware devices such as computers, servers, clients or hardware that has a connection with the network. In general, malware is interpreted as a program designed to damage a computer or server with malicious purposes such as gaining unauthorized access to certain systems by exploiting loopholes in a system or network. Most of today's malware is designed with many variations with different damage effects and this makes the ability to classify a similar variant of malware characteristics in a malware family is a good and also complex strategy in stopping malware. Research in this study attempts to classify malware using a dataset of malimg malware images which have a composition in the form of a grayscale bytemap with a total of 9,029 images of 25 different types of malware. By implementing the VGG-16 architecture and the comparative learning model, namely InceptionResNet-V2 in 2 different scenarios with scenario 1 using the original dataset and scenario 2 using the original dataset resulting from the undersampled process. Each scenario developed produces evaluation metrics values in the form of accuracy, precision, recall, and f1-score with the final result showing the highest score obtained in scenario 2 on the VGG-16 architecture with an accuracy score of 94.8% and the lowest in scenario 2 on the comparison architecture InceptionResNet-V2 with a score of 85.1%.