IVMCT: Image Visualization based Multiclass Malware Classification using Transfer Learning

Abstract

Computer systems have made it possible to transfer human life from the real world to virtual reality. This process has been accelerated by the Covid-19 virus. Cybercriminals have also switched from a real-life to a virtual one. Online, committing a crime is far easier than in real life. Cybercriminals often use malicious software (malware), to launch cyber-attacks. Apart from this polymorphic and metamorphic malware are used that use obfuscation techniques to create new malware variants. To effectively battle new malware types, you'll need to employ creative approaches that depart from the conventional. Traditionally signature-based techniques are used with machine learning algorithms to detect malware that is unable to catch its variants.  Deep learning (DL), which differs from typical machine learning methods, might be a potential approach to the challenge of identifying all varieties of malware. In the present study, an IVMCT framework is introduced which classifies malware using transfer learning. For this purpose, the MalImg dataset is used which is based on grayscale images converted from binaries of malware. The comparison of IVMCT is done with existing techniques which shows that our technique is better than existing techniques.