Research And Application Of Android Malware Detection Technology Based On Spark

With the development of the Mobile Internet and the popularization of Android devices,Android malicious software develops and spreads rapidly and posts great threat on the security of users' privacy and belongings.In the front of the rapid growth number of Android malicious software,the traditional methods which depend on manual statistics and analysis are not suitable for Android malicious software detection.Hence,it is of great significance to develop a new method which can detect Android malicious software more efficiently.To achieve these goals,this thesis focus on combining Data Mining and Spark cluster computing technique to propose a new kind of Android malicious detection method.The main research contents of this thesis are as follows:1.Proposing a new kind of Android feature based on Protected API which integrate permission and API information.To reduce the dimension of feature,a simple but effective method of feature selection named Double Threshold Filter which decrease the number of features automatically according to the distribution of features is proposed.2.Proposing an Android malware detection model based on Random Forests.A Random Forests classifier base on Spark is trained using features and feature selection mentioned above.The effectiveness of the model is verified by a plenty of experiments.3.Developing an Android malware detection system.An Android malware detection system is implemented using methods mentioned above.The detection system can classify massive Android applications automatically and effectively.These methods are verified by a large number of experiments and have good performance on detecting Android malware.Finally,an Android malware detection system with these techniques and methods is implemented successfully.