Research And Implementation Of Android Platform Application Behavior Analysis Based On Dynamic Detection

With the rapid development of mobile networking in recent years,the rapid spread of smart phones,especially the Android platform,smart phone market share increased year by year.Due to the Android platform's own characteristics and market characteristics,but also makes the current Android platform,malicious software to the user a huge loss.Therefore,Android platform software behavior research is the trend.For the PC and WEB software behavior research has become increasingly mature,and Android platform than the PC side in the software,hardware,there are differences,therefore,Android platform software behavior related research is necessary to carry out specialized research.At present,for the software behavior related research,has done a lot of work at home and abroad.There are methods based on feature code detection based on behavior detection.Based on the behavior of the method based on whether to run the application can be divided into static detection and dynamic detection.Static detection method is simple,more fooling way to identify,there are many drawbacks.Therefore,the main research dynamic detection,in the dynamic detection of the study,according to the perspective can be divided into application-level detection and system-level detection.Traditional application-level testing does not take into account the system environment factors,but also face some of the problems encountered in the code.The traditional system-level detection in most cases will change the system kernel,the system is unstable,and most of the research is to analyze the rules of the rules do not have the rules of machine learning,parameter optimization process.Therefore,this article is to start from the dynamic detection of the system to run the environmental data mining,without damaging the stability of the Android system kernel layer under the premise of the system environment to find hidden data behind the application behavior,and make monitoring With the continuous detection of the model can be achieved self-correction,and gradually improve the accuracy of recognition.The main work of this paper is as follows:1)to define the different software behaviors,to sample a large number of system environment data,and to cluster and quantify,to generate a set of characteristic sequences of single attribute data.2)to encode the data from multiple dimensions and generate the time series of the characteristics of the system environment data.3)the frequency of the symbol of different coding sequences is counted as the initial emission matrix of the hidden Markov model to model the hidden Markov model.4)the characteristic sequence of system environment data,using the system environment data to establish the HMM generated on the subsequent behavior of hidden Markov valuation,in order to achieve recognition for subsequent behavior,and in the subsequent recognition in the process of continuous optimization model.5,the method is proved to be effective by experiment.Therefore,through the comprehensive analysis of the system environment data to establish hidden Markov model for software behavior recognition than the traditional way to have some advantages,but also for Android platform software security research to provide basic research.