Font Size: a A A

Research On Android Malware Detection Based On Deep Learning

Posted on:2024-02-12Degree:MasterType:Thesis
Country:ChinaCandidate:M L ChenFull Text:PDF
GTID:2568307067972399Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The Android operating system is an open-source mobile terminal operating system.Therefore,it is one of the preferred targets of hacker attacks.And Android malware greatly threatens the privacy security and property security of mobile users.How to efficiently and accurately detect Android malware is one of the hot research directions in the field of information security.Therefore,this paper proposes an Android malware detection model based on deep learning algorithm,which can detect Android malware.The main research work is as follows:(1)An Android malware detection model based on deep learning is proposed.The model is divided into three stages,namely,feature extraction stage,feature vector quantification stage and model training stage.In the pre-processing stage,the four types of features are extracted and combined into a text document representing the Android application.In the feature vector quantification stage,the above text documents representing the Android application are processed into the form of digital vector.In the model training stage,the Bi LSTM network model can automatically learn the data information of Android malware.Increasing the number of hidden layers for the Bi LSTM network,which can improve the expression ability and generalization ability of the model.Secondly,a Dropout layer is added to prevent overfitting.(2)During the experimental analysis phase.First,the performance of the model under different size datasets was tested.Secondly,the detection effect of the model is tested under different classification characteristics.Then,compared with other algorithm models,the model detection accuracy in different data sets reached 98.08%and 96.15%,respectively,higher than other algorithm models.Finally,the trained network model is tested on my own data set.The experimental results were 92.77% and91.11% respectively.(3)The Android malware detection system is designed and implemented.The system can quickly determine whether the detected Android software is Android malware,and the system will display the extracted feature information and the final result of calculation on the detection result page.
Keywords/Search Tags:Android malware, Multifeature, Static analysis, Text classif ication, Deep learning
PDF Full Text Request
Related items