Research On Android Application Privacy Leakage Behavior Detection Method Based On Machine Learning

Smart phones have become an essential part of people ’s daily life.The surge in the number of Android applications has brought a lot of convenience to people,but also brought a series of problems,which poses a threat to the security of users ’sensitive data.Therefore,it is necessary to detect the privacy disclosure behavior of Android applications.The detection methods of Android application privacy leakage behavior are studied in this paper.Aiming at the problems of low detection rate and high false alarm rate of traditional static feature analysis,large overhead and low detection efficiency of dynamic detection methods,a privacy leakage behavior detection method for Android applications based on machine learning is proposed.The static and dynamic analysis are combined to improve the detection efficiency,and the permission comparison library is added to reduce the false alarm rate.The main research results are as follows :(1)Aiming at the problems of low detection rate of static analysis and low efficiency of dynamic analysis detection methods,an Android application privacy leakage behavior detection method based on multi-dimensional features is proposed.This method obtains the behavior information of the application by extracting sensitive permissions and API features,and adopts the Stacking algorithm model for training classification.(2)Aiming at the problem of high false alarm rate of static analysis permission features,an Android application privacy leakage behavior detection model based on permission comparison library is proposed.Using the mutual information model to select sensitive permission features,the application is divided into ten categories,and each category of applications sets different static permission sample libraries.The permission features of the application are compared with the comparison library to generate feature vectors.The Stacking model is also used to detect the privacy leakage behavior of Android applications.Experiments show that the detection model improves the shortcomings of the static detection proposed before,and further reduces the false alarm rate.(3)From the practical application,a privacy leakage behavior detection model for large-scale Android applications is proposed.