ANDROID MALWARE DETECTION USING IMAGE VISUALISATION AND MACHINE LEARNING

• Main Author: Panju Anandia, Degi
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53885
• Institution: Institut Teknologi Bandung

The popularity of Android smartphones has caused cyber criminals to develop malware on this platform. G DATA reports that there were more than 4.18 million Android malware in 2019 with an average of around 11,500 new Android malware appearing every day. Traditional malware detection techniques are no longer reliable to detect newly created malware in short period of time. In recent years, malware visualization techniques were introduced to detect malware. This technique is able to classify malware without the need for in-depth analysis. The stage of this technique is to change the classes.dex file in the apk file to a gray scale image and the image feature is extracted using the GIST descriptor. The image feature is then processed using machine learning to classify malware. Several studies have been carried out using this technique, but each researcher uses a private and dif erent dataset so that it cannot be concluded which method is the best. In this thesis, the author conducted an experiment to detect android malware using image visualization with publicly available datasets. The author uses three machine learning algorithms, namely k-nearest neighbor (KNN), Random Forest (RF), and Artificial Neural Network (ANN) to obtain the best performing algorithm. The experimental result shows that the RF algorithm produces the best performance with details of accuracy reaching 92.81%, precision 88.88%, and recall 83.72%. The time required to process the entire dataset consisting of 1,596 apk files is 43 minutes 16.69 seconds.