Droid permission miner: Mining prominent permissions for Android malware analysis

Abstract

In this paper, we propose static analysis of android malware files by mining prominent permissions. The proposed technique is implemented by extracting permissions from 436 .apk files. Feature pruning is carried out to investigate the impact of feature length on accuracy. The prominent features that give way to lesser misclassification are determined using Bi-Normal Separation (BNS) and Mutual Information (MI) feature selection techniques. Results suggest that Droid permission miner can be used for preliminary classification of Android package files.