**9. Experiments setup simulation and dataset**

The dataset was dealt with by Hacettepe University's Computer Engineering Multimedia Information Lab. It provides an RGB-based Core Fact Dataset for evaluating vision-based multiclass malware identification studies [6]. We used Keras Framework with TensorFlow in back-end, this about for deep learning libraries. For manipulate and process our images, we use Pandas and Scikit-leaning libraries. All experiments ran using Python 3.7 with notebook IDE.

**Figure 2.** *Malware sample from dataset.*

A Convolutional Neural Network (CNN, or ConvNet) is a sort of multilayer neural network designed to recognize visual patterns directly from pixel pictures with minimal pre-processing [7]. **Figure 2** shows an example of malware pictures. The ImageNet project is a massive graphic collection designed to be used in graphical object identification application testing. We utilized Keras, a deep-learning package. We simulate 18 types of malware using several deep transfer learning (DTL) models such as MobileNetV2, VGG16, and ResNet [8–12].
