Research And Implementation Of Android Malware Intelligent Detection Method

The rapid development of mobile Internet has promoted the popularization of mobile devices.Due to its good portability,Android operating system has become the largest intelligent operating system in the mobile application market at present.Subsequently,Android application as an important carrier of information communication has gradually penetrated into people's daily life.Because of its open source and popularity,Android has become the target of hackers and malware attackers,which may pose a serious threat to users such as privacy leakage,property security,and great losses to people.Therefore,it is of great research significance about how to detect Android malware effectively.By investigating existing Android malware detection methods,it is found that these methods mostly ignore structural semantic information of applications in the process of detecting malware.Although some Android malware detection methods capture structural semantic information of applications by extracting the graph structure of applications,these methods are very complicated and are not conducive to the rapid detection of Android malware.Therefore,based on the research on current Android malware detection methods,this paper proposes an Android malware detection method,which combines function call graph and graph kernel.This method captures structural semantic information of applications by extracting the function call graph.At the same time,it reduces the complexity of existing methods by using graph kernel.Experiments show that the method in this paper can effectively detect Android malware.(1)Based on the research on existing Android malware detection methods,this paper proposes an Android malware detection methods,which combines function call graph and graph kernel,which improves some existing Android malware detection approaches successfully.(2)Based on extracting function call graphs of applications,this paper automatically encodes these function call graphs through graph kernel,and then captures structural semantic information of applications.(3)In order to build the Android malware detection model quickly and effectively,based on the document embedding model,this paper proposes a graph embedding model,which is used to reduce the dimensionality of function call graphs.(4)To verify the effectiveness of the proposed approach in this paper,the detailed experiments are performed on the Genome and Drebin datasets.The experimental results show that the accuracy of the Android malware detection approach in this paper can reach 97.49% on the Genome dataset and 94.33% on the Drebin dataset.