Research On Similar Module Extraction For Android Malware And Applications

With the rapid development of the mobile,Android has become the most popular operating system on the mobile domain,and the huge number of users has attracted many developers to this field.As a result,the number of Android malware has contin-ued to increase,privacy and property security of mobile users are under a huge threat.Faced with a growing number of Android malware,traditional mobile malware detec-tion methods are no longer valid.Doing research on Android malware detection has practical value and significance.Effective and rapid implementation of Android malware detection and defense is a common goal of malware detection research at home and abroad,this thesis mainly includes Android malware similar module extraction method,deep learning-based de-tection model and system.1.Propose a similar module extraction method for Android malware:To deal with the problem of low accuracy and poor performance of existing detection methods,a key function-based Android malware similarity module extraction method is proposed,a dynamic self-growing similarity module database is constructed,and a subset of comparison samples is selected using the inverted indexing method to achieve rapid detection of Android malware.2.Propose an Android malware detection model based on deep learning:In order to extend the database of similar modules more effectively,a deep learning-based Android malware detection method is proposed,which extracts the three features of API sequence,permission and basic block of Android program samples,and integrates the XGBoost and convolutional neural network models for different features,greatly improving the accuracy of the detection model.3.Develop an Android malware detection system prototype:Build an Android mal-ware detection system with a self-growing similar module database as the core,integrating Android similar module extraction method and deep learning detec-tion model,as well as a variety of common mobile malware detection techniques,solving the problem of insufficient Android malware detection capability of tra-ditional systems and achieving high efficiency and accuracy of online real-time detectionThe research in this thesis focuses on Android malware detection and the construc-tion of Android similar module database,study the shortcomings of existing methods,and propose an Android malware detection framework with self-growing similar mod-ule database as the core,in which Android similar module extraction technique accuracy is above 91%on average,and the deep learning detection model has good performance under various evaluation indicators,and its AUC reaches above 95%.