| With the rapid development of mobile communication technology,applications on the mobile phone are changing with each passing day.As a result,more and more secrets are stored in the mobile phone,including personal,business,even state.However,when the mobile phone connects to WiFi,data leakage usually occurs,causing harm to national defense,business,as well as personal privacy.At present,data leakage from mobile terminals has become the research focus all over the world.The main data leakage source on the mobile phone is the large number of mobile phone applications.The application collects personal data from the mobile phone,regardless whether you know it or not.Once the application is infected with virus,data leakage occurs.In this paper,the basic principle and working technology of mobile Trojan virus is studied firstly.Then,various functional modules and performance are researched by analyzing the code of Loapi Trojan and kemoge virus.Trojan virus detection is an effective way to keep the app safe.Whereas,getting the virus signature is the critical step.Based on the above,this paper mainly research in the following two aspects:On one hand,optimizing the feature bank.Drebin data set is used here to extract the common permission in Trojan virus software samples,so as to form a feature library.By classifying and statistic,a large number of unrepresentative features are deleted.Simulation experiments are carried out on the optimized feature library.The result shows that,the optimized features can improve the detection rate and reduce the training time.On the other hand,proposing Trojan virus detection method based on BP neural network,and then improve it to GA-BP algorithm.By using the genetic algorithm,the power values and thresholds of BP’s neural network are optimized,avoiding problems such as the convergence of the standard BP algorithm is slow and easy to get to the local minimum.Then carry out simulation experiment on MATLAB to detect the Trojan horse virus,including network model building,network training and sample tests.The result shows that,the GA-BP algorithm has a shorter training time and higher detection rate. |
