AUTOMATED DETECTION OF PRIVACY RISKS IN ANDROID APPLICATIONS BASED ON DYNAMIC FEATURE SELECTION
Aimed at the user privacy leakage problem that may exist in Android applications,an automated detection model based on machine learning methods is proposed.This model chose to use the permission items applied by the App as features,dynamically selected the feature set,and used four classical machine learning algorithms to independently train and predict.And the most suitable privacy risk detection model for Android applications was determined.Experimental results show that the model can achieve an average prediction accuracy of more than 95％for privacy risk applications.This model can better manage application risk and protect user privacy from multiple aspects,which has high social benefit and practical value.
国家科技期刊平台
NSTL
万方数据
