Android Malicious Application Detection Model Based On Application Classification

Android plays an important role in the development of social intelligence.At the same time,malware attack based on the system has seriously threatened user privacy and system security.Qihoo 360 Internet Security Center monitored Android malware infections in 2018 by 110 million person-times.In 2017,Kaspersky Lab detected and intercepted a total of 5,730,916 Android malware installation packages.In order to improve the stability and security of Android and maintain the green and healthy development of the mobile application market,major Internet security companies and research institutes are still hard working.Android permission system plays an important role in the Android security model.The application requests software and hardware and user data resources through the permission system and achieves personalized features.In this thesis,we comprehensively researched Android security model and the recent development of malware detection research,and propose an Android malware detection model based on application classification,which mainly includes application classification module and category-based malicious detection module.The classification module takes the K-means algorithm as the core.By computing the similarity between the trusted application permission vectors in the training set,the trusted application is clustered into 32 classes and the clustering model is obtained.Then,the malicious application permission vector set is input into the model for classification.The detection module extracts the sensitive permission in each class as the input feature of the detection algorithm through the chi-square test,and trains the detection model.The experimental data contains 13,588 malicious applications and 13,382 trusted applications.The final experimental results show that the detection accuracy of the detection model can reach 94.02%.