| Due to the portability and powerful features of mobile smart terminals,people have increasingly used smart phones in recent years.Among them,Android mobile phones have become the smart phones with the highest market share due to the openness of their system platform and rich hardware.With the rapid growth of Android mobile phone sales,a large number of Android software came into being,which also includes a certain amount of malicious software.Malicious software causes malicious users to directly or indirectly cause economic loss or privacy leakage.In view of the unevenness of Android software on the market,how to distinguish between benign and malicious software has become a problem worthy of study.This dissertation aims to explore detection methods for Android malware.According to the characteristics of Android system mechanism and malicious software,a detection scheme combining static analysis and machine learning is proposed.Finally,a method for effectively distinguishing between benign and malicious software is implemented.The research process of this method mainly includes the following four aspects:(1)Based on the research and analysis of the architecture and security mechanisms of the Android system,permissions,sensitive API calls,Action attribute settings,and sensitive function pairs are used as malware classification features.(2)Download a large number of Android malicious and benign applications as experimental data sets,decompile the Android applications in the data set,and analyze the decompiled code to extract feature vectors.(3)Using the machine learning classification algorithm for classification detection,by comparing experiments,screening the most appropriate machine learning classification algorithm to improve the detection accuracy.According to the analysis of the experimental results,the classification accuracy is 96.2% with the K nearest neighbor classification algorithm,which has a good detection effect. |
