Graph Structure Oriented Android Malware Detection

The Android operating system is popular because of its advantages such as openness and portability,but its relatively simple security mechanism makes it the main target of many malware attacks.All kinds of security events bring serious losses to users,so it is of great importance to study effective malware detection methods in both academic and industrial fields.In this paper,a variety of typical Android malware detection methods and their respective advantages and disadvantages are studied.The emphasis is on the classification and comparison of malware detection methods based on deep learning.On this basis,by studying the related knowledge of Android system and malware analysis,the structure and training process of artificial neural network and convolution neural network,and combining the basic principle of the deep learning algorithm to deal with the image,a graph structure oriented Android malware detection method is proposed.The contents and results of this study are as follows.(1)The basic idea of the convolution neural network processing graph structure is introduced in detail,and the graph structure based Android malware detection framework is proposed,and the specific operation and related algorithms of each detection step are described.The function call relation of the application is expressed as a graph structure,and the standard graph structure input convolution neural network is obtained through the determination of the node sequence,the adjacency graph of the corresponding point and the normalization step.(2)In order to verify the performance of a graph oriented malware detection method,several groups of experiments include detection rate,detection efficiency,comparison with the comprehensive performance index of other models,learning curve and the influence of setting of threshold on the experimental results.Through the analysis of the experimental results,the method is proved to be feasible and the method has a certain degree of improvement in the detection efficiency and comprehensive performance compared with other methods.The factors that affect the experimental results are analyzed by the learning curve and the threshold setting.In this paper,based on the basic principle of deep learning algorithm processing images,the algorithm of convolution neural network processing graph structure is studied,and a graph structure oriented Android malware detection method is proposed,and the method is proved to be effective and feasible by various experiments.