Research And Implementation Of Android Malware Detection Based On AHP And Neural Network

Today,with the rapid development of science and technology,smartphone have become indispensable device in modern socity.Various applications hava emerged and offered lots of convenient service to people.The two major mobile operating system in the world are Android platform and Apple's iOS platform.Compared with iOS,the biggest feature of Android is its openness which offered developers huge amount of freedom to build application on Android device.In the other way,driven by enormous benefits of Android market,some hackers turn to set Android terminals as their attacking target due to its openness,the security of mobile terminal can't be ignored.Thus,the research of malware detection on Android platform is significant to the protection of Android ecosystem.For the research on Android malware detection methods,this paper introduces the research background,significance and development status of malware detection methods,and elaborate the history,architecture,and four major components of Android system.Then analyze the security mechanism of Android.In this paper,an Android malware detection method based on AHP and neural network is proposed in order to detect the APP's permission and sensitive API calling.The implementation of this method based on Client Server model,and the static detection is adopted as the detection method,that is to say,the detection is aiming to the installation package of Android application.The client mainly includes virus killing module,apk and app application selection module.The major server module is to generate malware detection rules set,first,malware samples decompile and get smali file,according to smali file extraction sensitive API,and then extract the API using AHP and improved BP neural network algorithm to combine weighting,the weighted sensitive API uses association rule mining algorithm to generate detection rules.Finally,through analyzing and contrasting the methods with a large number of experiments,it is verified that the method of this paper can effectively improve the accuracy of software detection and reduce the false alarm rate.