Architectural Design of Malware Infected File Detection Using Deep Learning

I. Abstract Currently, with the ongoing development of 5G and the IoT Cyberspace has grown to be a significant part of social and economic development thanks to the technology of social and economic development, as well as a fundamental domain of national security. Therefore, the detection of Malware and its variations are extremely important in cyberspace. Malware is a harmful software that is designed to damage and destroy computers and computer systems. Malware is constantly generating, updating, and manipulating which makes the detection more challenging. Viruses, Worms, Trojans, Ransomware, Adware, etc. are the common types of malware attacks. Signature based, Behavioral based and Heuristic ones are the three main methods used for malware detection. This paper discusses the architectural design for malware prediction in executable files of various applications and visualization techniques used to achieve better results by creating RGB images of these executable files and try to convert the malware prediction method into an image classification problem. InceptionV3 and the SVM model are used for feature extraction and prediction. Leopard mobile dataset and Malimg dataset are used here.