The Research Of Detection Technology Based On Android App Malicious Behavior

Mobile Internet has gradually replaced the traditional Internet evolution as the mainstream of the current era,under the background of this era,mobile intelligent terminal would replace the PC as a new information carrier.With the emergence of various levels of mobile application platform,for the whole of the mobile Internet provides millions of applications,these applications while enrich and improve people's lives,but also has provided people with a variety of potential safety problems,such as: electronic banking are stolen,personal privacy,malicious deduction,etc.These malicious behavior to a great extent,affect the normal operation of the intelligent mobile terminal applications market,because this area has been belong to the scope of big data,also need to perfect the relevant industry specifications,for this,based on the research of mobile application detects malicious act has the potential value and significance.Based on the study of domestic and foreign application detection method based on the Android platform,on the basis of combining with Android platform security mechanism and security architecture,component security,communications security,kernel security,in-depth study for the Android application level security threat,the comprehensive study on the threat detection and model building.In many categories in the App to select three App as sample space,the inverse analysis was carried out on the sample App,extraction and application permissions,broadcast messages,API calls,such as related characteristics,according to these features to generate a binary vector,and then based on vector selection feature subset,the optimization algorithm based on support vector machine,the random forest optimization algorithm,enhanced adaptive optimization algorithm of three kinds of machine learning algorithms to classify the feature subset,constructing classification model,put forward a kind of Android applications of malicious behavior detection based on machine learning model,using the above technology structures,experimental environment,the relevant classification samples are provided based on performance of the algorithm is achieved,to complete the real-time detection of all kinds of malicious behaviour.