| The development and popularization of communication technology and Internet of things technology provides conditions for the emergence of new things,and the demand for real-time processing also arises with it.Applications such as autonomous driving,virtual reality(VR),and augmented reality(AR)need to get the calculation results in milliseconds.Due to the limitation of geographical location,the traditional cloud computing mode cannot meet the needs of real-time tasks in terms of time delay.Therefore,the concept of Edge Computing is proposed,which sinks computing resources to the edge of the network.Users can offload computing tasks to the nearest edge server for processing,which can shorten the data transmission delay and improve the service quality to a certain extent.However,edge computing offloading still faces some challenges.Computing resources at the network edge are limited,and the performance between different edge servers is different.Therefore,different offloading strategies have a great impact on improving the quality of service.How to design an efficient offloading strategy has become the primary issue of edge computing.In addition,the offloading data contains a lot of user privacy.How to ensure the security of user privacy has become a problem that cannot be ignored in the field of edge computing.This thesis focuses on the above problems and conducts research on edge computing offloading strategies.The main work of this thesis is as follows:First of all,we have introduced the relevant theories of edge computing offloading,including application fields,system architecture,and offloading performance indicators,expounded the concept of fine-grained offloading,and analyzed the common security threats and privacy protection methods in the offloading process.In the next part,in order to improve the quality of service,taking the multi-terminal multi-server as the offloading scenario,we have proposed an edge computing offloading strategy based on fine-grained.At present,the most of edge computing offloading takes the overall task as the offloading unit.In this thesis,the overall task is divided into several sub-tasks with dependencies,which can increase the possibility of parallel computing within the task,thereby further shortening the processing delay of edge computing.Moreover,on the basis of refining the task granularity,the offloading model fully considers the factors of terminal energy consumption,user cost and load balance,which improves the quality of service from multiple perspectives.The above offloading model is solved by the nonlinear updating particle swarm optimization algorithm.The simulation results show that the edge computing offloading strategy based on fine-grained performance is better than the overall offloading in various indicators.In the end,we have proposed an edge computing offloading strategy based on privacy protection for the privacy security issues in the offloading process.Using k-anonymity technology,the quasi-identification information in the offloading data is hidden by establishing the generalization tree,ensuring that there are k pieces of data in each equivalence class,cutting off the connection between users and privacy information,and reducing the probability of malicious attackers locking users.At the same time,privacy entropy is introduced into the offloading model to increase the confusion of data distribution on different edge servers,thereby improving the security of private data during offloading.We use the Kullback-Leibler divergence evaluation index to measure the difference between the offloading strategy and the random offloading distribution,which can further measure the privacy protection effect of the offloading model.The offloading model is solved by deep reinforcement learning,and the simulation results show that the privacy protection offloading model not only control the time delay and energy consumption,but also realizes the user’s privacy protection to a certain extent,which verifies the effectiveness of the edge computing offloading strategy based on privacy protection. |
PDF Full Text Request