Android Malware Detection Method Based On Function Call Graph

As an open-source mobile operating system based on Linux kernel,Android is getting more and more popular in recent years.The openness of Android system is a double-edged sword.In one hand,it offers convenience for developers and end-users.In the other hand,it caused the vulnerability of Android system which leads to the generation of a large number of malicious applications.The malware makes the ecological environment of Android operating system getting worse and worse,many users are impacted by it.The research on Android security mechanism and malicious application detection becomes an important part of mobile security under such a circumstance.This paper investigates the malware detection algorithm based on function call graph of Android application.The existing research work has shown that feature engineering is important in improving the performance of Android malware detection model.To a certain extent,the feature engineering is more important than the selection of detection algorithms.First,the deficiency of current malware detection models is analyzed,then a malware feature extraction algorithm based on graph convolution network is proposed.The feature extraction algorithm effectively improves the performance of Android malware detection model.The contributions of this paper is listed as follows:(1)Summarized the problems of related research at home and abroad.That is,the static features used in the existing research,such as permissions,can't reflect the behavior characteristics of Android applications,and the dynamic analysis tools also have certain limitations.(2)For malware detection problems,the behavioral characteristics of software and other static features are equally critical.Given the limitations and complexity of dynamic analysis tools,the importance of constructing behavioral characteristics from static analysis of Android software is significant.This paper innovatively extracts the behavioral feature representation of Android software from the static feature-function call graph,and studied the Android software feature extraction algorithm based on graph convolution network which improves the accuracy of current malware detection model.(3)Implemented and evaluated the above Android malware detection model and system.The results of experiments show that proposed model has certain advancement compared with the traditional model and methods.This paper mainly studies the static feature extraction algorithm of Android applications.The results of experiments proved the proposed method has effectively improved the accuracy and efficiency of current malware detection technology.