Review on Malware Classification and Malware Detection Using Transfer Learning Approach

Abstract

The amount of malware infecting computers and other communication devices and migrating across the internet has greatly increased over the years. Numerous methods and procedures have been put forth up to this point to find and eliminate these hostile agents. However, a lot of malwares is still being developed, which can get past some cutting-edge malware detection algorithms, as new and automated malware production techniques emerge. Consequently, there is a need for the classification and detection of these antagonistic agents that have the potential to compromise the security of individuals, businesses, and a wide range of other digital assets. To develop a new improved approach for efficient zero-day malware detection, there is a compelling need to reduce bias and objectively assess these methods. This study focuses on investigating transfer learning strategies for malware picture classification, including AlexNet, VGG16, VGG19, GoogLeNet, and ResNet. Here, malware binaries are turned into grayscale images before being processed using models for classification based on transfer learning. As transfer learning uses pre-trained models, the primary objective is to save training time. In addition, an effective model for malware classification will be built in order to obtain performance that is unbiased.