The Research On Malware Detection Technology Of Android Based On Image Texture And BP Neural Network

With the rapid development of mobile Internet,Android system with open source property occupies more than 85 percent of the global market.However,due to the open and fragment features of Android,and the popularity of Android devices,many hackers will think of ways to benefit from this mass of unsophisticated users,which makes the quantity of Android mobile malware more and more,and seriously plagued every mobile user's privacy and safety.There the design of a better Android malware detection method is imminent.This paper focuses on the detection of malicious software based on image texture and BP neural network.The main work of this paper is as follows:(1)This paper firstly analyzes the architecture of Android system,four components,ART and Dalvik virtual machine,and their security models.The paper analyzes the sandboxing mechanism,permissions mechanism and digital signatures.At the same time,the malicious software at tacking attacking technology and detecting technology of Android system are analyzed in detail.By collecting large quantities of Android malwares,we summarizes the source of security threats under the android platform,and malware classification.Then for the quick and effective detection analysis of Android malware,the paper summarizes the common detection techniques,including static analysis and dynamic analysis technology,and popular machine-learning-based detection technology and virus visualization analysis technology.(2)On the basis of this,the traditional malware detection method has the problems of slow detection speed,weak adaptability of multiple platform detection,etc.This paper propose a malicious software detection method based on image texture and BP neural network algorithm.By combining the technology of image analysis and the malicious software detection,the malicious software is converted into gray-scale image.GLCM(Gray-Level Co-Occurrence Matrix)algorithm and GIST(Generalized Search Trees)algorithm are used to extract the texture features and BP neural network algorithm is used for learning and training to rapidly detect malicious software from different platforms.(3)Based on the method proposed in this paper,we realized prototype of image texture extraction and detection of malicious software,the system can be done fast,multiple platform testing adaptable and high accuracy characteristics.