Investigation of Feature Selection Methods for Android Malware Analysis

AbstractIn this paper we present a method for detecting malicious Android applications using feature selection methods. Three distinguishing features i.e. opcodes, methods and strings are extracted from each Android file and using feature selection techniques, prominent and diverse, top ranking features are mined. Different tree classifiers are used to categorize Android files as either malware or benign. Results show that methods is the most credible feature, which gives accuracy of 88.75% with 600 attributes using Correlation Feature Selection method and Adaboost with J48 as base classifier