Android Application Security Detection Method Based on Metropolis Algorithm

Abstract

Smartphones are becoming more and more popular due to their convenience, Android system smartphones accounting for most markets. The security of Android applications (Apps) has become especially important. Currently, there are many security detection methods based on Android Apps. But these methods rarely consider the uncertainty of the feature. In our paper, an Android applications security detection method based on Metropolis algorithm is designed. This method analyzes Android’s 24 dangerous permissions using the Metropolis algorithm, removes uncertainty permissions, and extracts certain permission features. Then, the features are learned and classified through the classification technology of machine learning. Our approach uses 1501 Android Apps samples (including 870 benign samples and 631 malicious samples). From the experiment, our method reduces the detection features, and the accuracy of malicious application detection can reach 93.5%.