Research On Privacy Protection Offloading Scheme In Edge Computing

In the context of the new era of Internet of things,new scenarios such as smart home,digital wisdom medical,autonomous driving and smart city are emerging rapidly,followed by the increasing demand for computing power and explosive growth of terminal data scale.This requires more computing power,storage capacity,battery life and better performance of the central network.In order to meet these requirements,the edge computing emerges as the times require.Its core function is to sink the functions and services that originally belong to the cloud center to the edge network,which is very important for the realization of low latency,high bandwidth and flexible services.However,the edge computing mode brings a variety of interaction modes and new use experience for terminal devices and users,and at the same time,it also brings new privacy and security issues to users.In the edge computing,the terminal needs to offload the computing tasks and data to the edge server to obtain the corresponding services.However,in this open service delivery mode,it is difficult to completely eliminate the existence of malicious nodes in the edge system.These malicious nodes seriously threaten the security of the system.Besides malicious nodes,honest and curious edge service nodes are also worthy of our attentions.These untrusted nodes collect and analyze users' information,such as identity information,sensitive data,location information and background knowledge,from the data packets uploaded by the user terminal and the task unloading log,which will threaten the privacy of users.At the same time,due to the limited hardware resources of terminal devices,it is difficult to implement complex security encryption algorithms,which makes the privacy security problem in edge computing mode more serious.In this thesis,we focus on privacy protection methods and edge offloading models in edge computing system.Firstly,aiming at the multi-user binary offloading mode in edge computing,we propose a multi-user anonymous binary offloading scheme based on the quadratic programming.Then,we propose a privacy aware offloading scheme based on the Markov decision for single user multi-application data offloading scenario.The main work is as follows: 1.Aiming at the problem that malicious terminals or untrusted nodes invade the edge system caused by the openness of the edge system,an anonymous identity authentication mechanism is designed to enhance the security of the system and the privacy of the terminal.Then,aiming at the establishment of multi-user binary offloading mode and the solution of optimal offloading decision-making in edge computing,a multi-user binary unloading scheme based on the quadratic programming is designed.In this scheme,the unloading decision-making of the edge system is modeled as a non-convex quadratic constrained quadratic programming problem.For the NP hard problem in the model,the semi-definite relaxation method is used to obtain the approximate optimal solution,and then a low complexity decision extraction method is designed to extract the final decision.In particular,in the process of solving the model,a step-by-step iterative method of unloading decision and resource allocation is proposed,which reduces the algorithm complexity.Finally,the experimental results show that compared with the benchmark scheme,our scheme still has better system performances while ensuring users' privacy.2.Aiming at the single user multi-application data unloading process in the intelligent medical Internet of things scenario and the privacy security problems caused by the unloading,a privacy aware offloading scheme based on the Markov decision is proposed,which decomposes the unloading process into time slots and models it into a Markov decision process.After that,the privacy function is added to the value function of the system to optimize the unloading strategy,so as to protect the privacy of terminal devices.In the process of obtaining the optimal policy,aiming at the problem that the unloading data size in the state space is a complex continuous function and the system state transition model is unknown,we use the model-free DDQN algorithm to learn the system environment and get the optimal unloading strategy,thus minimize the terminal delay and energy consumptions on the basis of ensuring users' privacy.In particular,we improve the greedy strategy in the DDQN algorithm learning process.The experimental results show that the convergence results are better than those of the benchmark schemes.