A Preliminary Study on Malware Classification using Image Pattern

Abstract

Android operating system occupies more than 80% of the world market share in mobile operating system. The popularity of the Android operating system motivates cybercriminals to develop malware targeting this platform. In the first half of the year 2021, there were 1.3 million new malicious Android applications circulated on the globe which the malware analysts need to analyse. Traditional malware analysis techniques are no longer reliable to analyse the huge amount of malware, and they require more resources to process and store them. This research proposed a different approach to analyse Android malware and maintain high classification accuracy with minimal resource usage. 3,900 Android applications consist of malware downloaded from Android Malware Dataset, and benign samples downloaded from APKMirror website were used in this research.  The preliminary results of the study show that the image pattern from the same family are analogous meanwhile different family of malware presents distinctive image pattern. Thus, further analysis is needed for different sizes and rotation of extracted malware images.