Design And Implement Of Android Malware Detection System Based On Deep Learning

With the rapid improvement in the number of mobile devices and the market share of the Android operating system,the security of application under the Android platform has received more and more attention and challenges.With the development of machine learning technology and deep learning technology,how to apply related technology to efficient and accurate malware detection has become the research direction.The main purpose of this paper is to design and implement a detection system based on deep learning technology,which can detect and identify application under Android platform.This paper uses the sequence of operational instructions obtained by static decompilation and the sequence of sensitive behavior detected by dynamic analysis as features.Referring to the neural network model in the field of natural language processing,the operation instruction sequence is regarded as a natural language of the machine.Finally,a hybrid neural network model based on convolutional neural network and Long Short-Term Memory network is built to train.The neural network model is implemented and tested under the TensorFlow platform,enabling detection and identification of malicious application.This model performs well on the collected data sets,and the experimental results can reach 91.3% accuracy.In addition,an Android malware detection system with this neural network model as the core detection module,combined with blacklist mechanism,decompilation analysis,simulation running detection and other modules is designed and implemented.The system has the characteristics of comprehensive coverage,high accuracy,fast operation efficiency and good expandability,and can detect and identify most of malware.