February 2016
·
26 Reads
To accurately know the contributions of a single feature and a single data mining algorithm to high detection accuracy for malware detection, this paper puts forward a mobile malware detection approach using ensemble techniques for the Android platform. The proposed approach extracts three kinds of features from a given mobile application, including privilege feature, component feature and API call feature. Several classification models are built for each kind of feature using several base classifiers respectively. A consensus function for each feature is designed to make decision to obtain an optimal classification output. In the next step, another consensus function is designed and applied to the outputs from all kinds of features in order to obtain the final classification output. This paper carries out the empirical experiment evaluation on mobile applications from the real world application markets, and the compared results show that our approach can get a better detection accuracy in terms of F1 score than a single data mining algorithm. © 2016, Editorial Department of Journal of Nanjing University of Science and Technology. All right reserved.