Improved Malware Detection Results using Visualization-Based Detection Techniques ant Convolutional Neural Network

The rapid advancement of internet technology has carried major changes in the use of software, st the industry growing up and competing with each other to present innovations for making life easier. However, software spread on various platforms contains a lot of malware, which can compromise the security of users’ personal information. Until now, researchers continue to try various methods of malware detection to minimize the weaknesses of dynamic and static methods. This study will compare the malware detection method using Visualization-Based Detection Techniques and one type of deep learning, which is Convolutional Neural Network (CNN). The evaluation experiment used two methods of image augmentation, which are the B2IMG dan Gabor Filter methods. The CNN architectures that we used, such as VGG3 dan other architectures combined with several convolutional layers and pooling layers. The evaluation run result indicates that the B2IMG visualization method got a maximum accuracy value of 99,86% and an F-score of 97,00% when using colored RGB image format.