Research On Android Malware Detection Method Based On Opcode Features

With the rapid development of mobile Internet technology,more and more smart devices have entered people's lives,and the smart phone industry has also emerged.Because Android system occupies a large market share in the global mobile operating system market,and because of its open source nature,the Android platform is vulnerable to malware.The proliferation of Android malware has brought huge economic losses to users,and it has also had a bad impact on the future of Android.Therefore,research on Android malware detection methods has become a hot research topic.In this paper,the Android platform architecture,Word2 Vec word vector training model and convolutional neural network are analyzed in detail,which lays a theoretical foundation for the subsequent proposed detection model.Then,this paper proposed an Android malware detection model based on opcode features.Firstly,the model extracted the opcode sequence feature of the Android application.Then,Using the opcode sequence trained the Word2 Vec model to obtain the opcode word vector and constructed the opcode feature matrix.Secondly,using the opcode feature matrix trained the convolutional neural network to realize the detection and classification of malware.The experimental results show that the detection model has better malware detection effect.Finally,aiming at the deficiencies of the detection model,a simplified scheme of opcode instruction classification was proposed.This scheme was used to simplify and extend the opcode feature by constructing the sequence of instruction function features.Then,The Dualchannel convolutional neural network was designed to integrates the opcode features and instruction function features.Contrastive experiments show that the simplified classification scheme of the opcode can effectively improve the detection effect of the model.Compared with the detection effects of other models,the model also has certain advantages.