Research On Android Terminals Control Technology Based On Indoor And Outdoor Scenes Recognition

With the advent of the mobile Internet era,Android mobile terminals are becoming more and more popular in people's daily life and work.However,with the increasing power,mobile phones often become the source of privacy leakage within a company.What's more,the work efficiency of core staff can be decreased by unmanageable use of mobile phones.In order to solve the above problems,a viable and real-time management program is needed to control the employees' mobile phones usage.Taking into account that employees' handsets should be controlled more strictly in the office environment,different control strategies should be adopted for indoor and outdoor scenes.Therefore,an accurate and efficient indoor and outdoor scenes recognition technology is required as a basis for the upper management and control application.Based on the above consideration,the research objective of this thesis is to design an indoor and outdoor scenes classification scheme with high accuracy,high performance and low latency,as well as real-time management and control technologies that can impose fine-grained restrictions on the specific functions of Android terminals.The Indoor and outdoor scenes recognition scheme is aimed at the recognition of both indoor and outdoor conditions.The main content of this thesis is as follows:(1)Research on the indoor and outdoor scenes recognition technology.To recognize indoor and outdoor scenes,the characteristics of Android terminals' lightweight sensors and wireless signals,which are different in indoor and outdoor environments,are used.Based on the research of previous schemes,the classification effects of many common machine learning algorithms are compared through supervised training processes.Decision tree can be found more applicable to current problems from the comparison result.Then,a self-learning decision tree(SLDT)algorithm based on semi-supervised training is proposed.High-confidence unlabeled samples are used to update the classification model continuously through a semisupervised decision tree training method in the algorithm.SLDT algorithm is a solution to some of deficiencies of the previous schemes and decreased accuracy of the classification model under a new environment.(2)Research on Android terminals management and control technology.The custom code with the function of management and control is injected into the system service process by Inject and Hook technology in this thesis.Current requested services can be found by hijacking the address of function ioctl,which is a key function in inter-process communication(IPC).After that,IPC data packets can be monitored and parsed by the injected code.If the requested services are in the blacklist of control strategy,the relevant data packets are modified to realize real-time management and control.(3)With the help of the previous research,a prototype system based on indoor and outdoor scenes recognition for Android terminals control is designed and implemented.In the system,different control strategies for the monitored Android terminals can be adjusted automatically according to terminals' current scenarios.It can be showed that the system has good compatibility and outstanding characteristics of high performance,low energy consumption and low latency in the tests.