Research On Privacy Protection Offloading Algorithm In Edge Computing Scenarios

With the popularity of 5G commercial and Internet of Things(Io T)devices,as well as the everincreasing scale of user mobile data,the new architecture of mobile edge computing(MEC)has become a research hotspot.Computing offloading is one of the key technologies of edge computing.An efficient and safe offloading strategy is the basis for improving user experience.However,the computing offloading strategy in edge computing scenarios still faces a series of difficulties and challenges,such as the real-time performance of data transmission,the efficient execution of computing tasks,the security of personal privacy,and the effectiveness of equipment energy utilization.Existing work has limited research on the security of the offloading strategy in MEC.This thesis focuses on scientific issues such as efficient transmission,real-time communication,resource allocation,and user privacy security in MEC,and conducts in-depth research on the privacy protection offloading algorithm in MEC.Firstly,it analyzes the data security and privacy protection issues in the MEC system,and introduces the core privacy protection technologies for data security and common anonymity models.In addition,the principles of commonly used machine learning algorithms and their respective advantages and disadvantages for calculating unloading strategies are explained.Secondly,in view of the privacy leakage problem caused by the frequency of user computing task offloading in MEC,a privacy protection offloading algorithm based on virtual mapping is proposed.Privacy constraints are introduced into the deep reinforcement learning algorithm generated by traditional offloading strategies.A reasonable virtual task mapping mechanism is formulated according to the user’s offloading characteristics to reduce the amount of privacy contained in the computing task.At the same time,the offloading policy is dynamically adjusted to reduce the amount of accumulated privacy of users on the MEC server and the risk of user privacy leakage.The simulation results show that,compared with the traditional offloading algorithm,the proposed privacy protection offloading algorithm can effectively protect user privacy while maintaining high offloading performance.Thirdly,from the perspective of multi-users,k-anonymity technology is introduced into the traditional offloading strategy.By anonymizing multiple users in the same time slot to calculate the offload probability of tasks,the target users have the same offloading feature with at least k-1 other users which can avoid the identification of attackers.The generalized offloading probability of users in the equivalent group effectively reduces the amount of privacy contained in the computing task and reduces the risk of user privacy leakage.There is no need to formulate different privacy protection strategies for each user’s offloading feature.For multi-user scenarios,the algorithm complexity is effectively reduced,and the calculation and storage pressure of the server is also reduced.The simulation results show that compared with the privacy protection offloading method based on virtual mapping,this scheme has better offloading performance under the same privacy protection effect,and is more suitable for multi-user scenarios.