Overprivileged Permission Detection For Android Applications

Smartphone plays an indispensable and important part in modern human life,bringing convenience and richness to user's life.The multiple functions provided by the smartphone are based on the various applications installed on it,hence the development of smartphones leads to a great boom of application markets.As the important private item of user,there are more or less personal information in the smartphone,such as photos,videos,short messages,contacts and so on.Therefore,the privacy security of smartphones has received lots of attention.Android defines permissions related to user privacy as dangerous level and adopts permission mechanism to restrict applications from accessing important resources of the smartphone.However,the security concerns caused by the overprivileged permission are gradually increasing due to the absence of strict supervision mechanism in application markets and the lack of strict development specifications of developers.The overprivileged permission means the extra permission declared by the application but has nothing to do with its function.Most users grant any permission declared by the application since they know little about permissions,which allows the application to collect privacy data having no relation with its function,intensifying the risk of privacy leakage.Aiming at this problem and on the basis of the analysis of Android permission mechanism,in this thesis we study the relation between the dangerous permissions and the categories the application belongs to.Our research is based on the frequent item set mining.By applying the FP-growth algorithm to many popular applications of Google Play,we get the maximum frequent permission set of each category.The permissions in each set represent those permissions related to the functionality of the applications under the category.We design and implement a detection system for overprivileged permission,which is named Droidtector(Android detector).The system is developed in Python and can operate in online or offline mode.The online mode needs PC to access the Internet while the offline mode needs no connection to the Internet but user should have got the APK file and known its category before detection,user can choose any mode according to their situation.The results of effectiveness verification confirm the correctness of the detection results of Droidtector.We run Droidtector on 1000 popular applications crawled from Google Play and find 479 of them contain overprivileged permission,accounting for about 48% of all the sample applications.By comparing our testing results with that of previous studies,we can see that the proportion of overprivileged applications are increasing on the whole,indicating that the phenomenon of permission overprivilege is becoming serious year by year.