PMMSA: Security analysis system for android wearable applications based on permission matching and malware similarity analysis

Abstract

Wearable devices based on the Android system are developing rapidly, but the research on their application security is still lacking. Therefore, this paper designs an Android wearable application security analysis system-PMMSA. PMMSA first conducts permission matching analysis of smartphone and wearable device applications to ensure the safety of application installation. Secondly, it performs malicious application similarity analysis on wearable endpoint applications to ensure the security of application usage. In the study of malicious application similarity analysis, due to the small number of Android wearable applications, this paper proposes a binary adjacency (BA) oversampling method to expand the number of applications. In addition, we propose the C-M-KNN model to compare the similarity of wearable devices and malicious applications, which uses KNN as the base analysis method. To reduce the detection time, we introduce the mean center strategy. We also introduce convolutional neural networks to improve the accuracy of MC strategies. The experimental results on Google Play Store and VirusShare datasets show that 41 apps have permission mismatch, and the false positive rate of benign samples of Android wearable apps is 1.55%.