Özet:

The internet made an individuals life very easy and more productive, but there are some associated threats linked to the internet and devices. Malware is considered the most severe threat for decades to the digital world and malware variants identification and classification is the most vital and critical research problem. It is an invasive malicious code that accesses devices, information, and services without the permission, knowledge of the user. Researchers, analysts and antivirus companies are incessantly inventing and implementing new strategies to fight back malware and its variants. In the last decade, one of the strategies is extensively used in the field of malware detection and classification is the deep learning methods using malware visualization. Results revealed that using visualization; malware can be identified, classified more promptly, efficiently, and accurately. Deep learning algorithms vary according to applications, architecture, and uses, so it is required to review and inspect the work based on deep learning to use malware visualization to know the recent approaches and innovations that have been established, to identify problems, current issues, challenges, and of course at the same time to motivate potential research directions. In this effort, an extensive survey of works that utilized deep learning methods using malware image representation, for malware classification is reviewed with a detailed discussion on key methods such as data sets description, malware image representation strategies, and deep learning architectures of parameters, contributions, and limitations. A comparison of the reviewed work is presented based on various key factors.

Anahtar Kelimeler: