Permission Weighting Approaches in Permission-Based Android Malware Detection

Abstract

With the increasing use of mobile devices in daily life, the number of malware running on mobile devices is increasing. Increased malware may cause material and non-pecuniary damage, such as the seizure of personal information of users or the deterioration of personal data. Therefore, the need for systems that detect malware with high accuracy is increasing day by day. In this study, it is aimed to determine malware using the machine learning based static analysis technique for Android operating systems. In order to obtain high performance rates in malware detection, 14 different terms weighting techniques frequently used in text classification have been extensively adapted to this. Adapted methods were tested on 2 different datasets and compared with 3 different classification algorithms. The most successful classification result on the AMD data set was obtained from binary term weighting technique and support vector machine classification algorithm. The most successful classification result on the MODROID data set was obtained from discriminative weighting technique and support vector machine classification algorithm.