Research On High Confidence Resource Allocation Technology In Edge Computing

With the advent of 5G and the rapid development of Internet of Things,wearable devices,face recognition,and various new terminals have emerged continuously.Because of the limited computing capabilities and energy in new terminals,traditional cloud computing with high latency and high network bandwidth usage cannot meet the needs of the era.To relieve network load and cloud computing pressure,edge computing and cloud-edge computing architecture are widely adopted.Edge computing can provide resources to improve user experience,the research on task offloading and resource allocation optimization is currently a hot issue.Moreover,the open environment is prone to cause the issues of device entity trust and data privacy leakage,the research on trust mechanisms is also a hotspot in the edge computing field.To solve the resource allocation and security issues of edge computing,the paper adopts a data privacy and entity trust mechanism in the resource-limited and changing environment.For the dynamic scenarios of multi-user multi-cell full offloading and multiuser single-cell partial offloading,adaptive optimization algorithms for task offloading and resource allocation in a high-confidence environment are respectively proposed,which effectively improve user experience and security.The innovations of the paper mainly include the following aspects:(1)For the problem of distrust between the user terminals and edge servers in the multi-cell multi-user scenario,the paper constructs a lightweight trust mechanism and introduces the varied trust coefficient as an incentive to enable the users quickly distinguish malicious servers.Considering the limited and dynamic computing resources,the problem of complete offloading of tasks is transformed into an allocation problem of computing resources to minimize the total energy consumption.To handle the problems of explosion of action space and difficulty in convergence of centralized single-agent algorithms with the increasing of users,the system is established as a Markov Game model and a distributed resource allocation algorithm based on trust model and multiagent(DRATMMA)with the centralized training and distributed implementation is proposed.The simulation results show that the algorithm can effectively resist malicious attacks,and can make the reasonable offloading decisions based on different states,reducing the total energy consumption of the system and providing a better user experience.(2)To solve the privacy leakage problem caused by eavesdropping under the offloading process in the single-cell multi-user scenario,physical layer security is adopted to make confidentiality measures according to the security requirement of user terminals.Because of the dynamic wireless channel state and the computing resources required by users,the tasks partial offloading problem is modeled as a joint optimization problem of computing resources and power under energy constraint and confidentiality to minimize the average processing delay,meanwhile the system is modeled as a Markov Decision Process,and a resource allocation algorithm based on physical layer security and deep deterministic policy gradient(RAPLSDDPG)is proposed.The simulation results show that the algorithm can approach the optimal performance,has lower computational complexity and adaptability,and can make a better offloading and resource allocation strategy in the trusted environment,effectively reducing the average processing delay and improving the robustness of system.Finally,the main content and achievements of this paper are systematically summarized and the areas that can be optimized and improved are also prospected.This paper includes 25 figures,7 tables and 48 references.