| With the arrival of the existing computing-intensive intelligent applications and the fifth-generation communication technology(5G)era,70%of the rapidly growing data at the edge of the Internet of Things(IoT)will be processed at the network edge.Cloud computing technology has been unable to meet the requirements of low energy consumption,low service latency,localization and high bandwidth.In order to process a large amount of data at the network edge in time,edge computing paradigm has been developing rapidly in recent years.The server is deployed near the data source to expand the computing power to the network edge so as to be closer to the end users,which greatly reduces the network transmission delay and improves the network service experience of the end users.Facing the challenges of diversity task requirements and limited computing capacity of terminal devices,computing offloading collaboration technology will effectively make use of the horizontal computing collaboration between idle device nodes,the vertical collaboration between terminal device node and edge server,which can achieve better resource utilization,balance the computing load,further obtain low latency and improve the perceived performance of users.However,most of the schemes for offloading task computation in edge computing environments with the goal of minimizing latency or energy consumption do not consider the service reliability of the service resource nodes.When the malicious node attack behavior increases in the network service environment,it will destroy the stability of the network service environment and reduce the network efficiency.Meanwhile,in the process of computing offloading,the traditional optimization algorithm model is solidified,which can no longer meet the needs of computing offloading in the edge environment.There still exists several challenges behind the edge computing environment:the internal attack of malicious nodes in the trust mechanism;the problem of user preferences and task requirements;the problem of task offloading optimization,etc.This paper focuses on the issues related to trusted resource allocation and task offloading in the edge computing environment,and introduces the writing routes of edge computing network architecture,trust evaluation management technology,task computing offloading technology and deep reinforcement learning technology respectively,so as to optimize the computing efficiency,computing delay and energy consumption of the system.The main work and contributions of this paper are as follows.(1)Aiming at the internal attack of malicious nodes in trust mechanism for edge computing environment,a multi-source trust mechanism based on time attenuation factor is proposed.In the edge computing environment,numerous terminal devices and edge servers are involved with providing service computing and data transportation,which makes the network service environment vulnerable to attacks.In order to solve the problem of malicious node attack,a trust evaluation mechanism of node interaction behavior is introduced.First,trust relationships among edge nodes are established in an open edge computing environment,and a multi-source trust fusion algorithm based on time decay is proposed,which aggregates direct interaction trust based on time decay factor,direct recommendation trust based on reliable feedback with interaction of evaluation node and indirect recommendation trust based on Jaccard similarity without interaction of evaluation node to calculate the global trust of target nodes.Then,the time influence factor is introduced to describe the rationality of trust,to improve the accuracy and maintain the freshness of trust.At the same time,the introduction of reward and punishment factors to stimulate benign node interaction and dynamically updating the trust value can effectively prevent swing attacks.Finally,The simulation experiments show that compared with the existing trust model,the proposed algorithm promotes the computational efficiency and interaction success rate by 29.25%and 12%,respectively.(2)Aiming at the problems of user preference and task demand in edge computing environment,a resource allocation strategy based on multi-feedback trust mechanism is proposed.Facing the diversified tasks and massive,dynamic and heterogeneous resources in edge computing environments,how to obtain reliable and fast-responding services and assign application tasks to resource nodes that meet task requirements and user preferences for execution is a challenge.IoT edge computing environment faces different types of serious attacks,e.g.,bad-mouthing attacks,false attacks,and collusion attacks,etc.Therefore,trust evaluation among edge nodes is necessary for users to provide a reliable service environment.However,existing trust computing schemes have long response cycles and low malicious detection rates in dynamic environments.To solve these problems,a resource allocation strategy based on a multi-feedback trust mechanism is proposed.Firstly,a reliable and efficient multi-feedback trust resource allocation framework is established,which effectively improves the security and operational efficiency of trust computing model and task computing environment.Secondly,according to the dynamic monitoring of nodes by broker,a multi-feedback trust aggregation model based on time decay and interaction frequency is proposed to provide the running environment of trusted services.Next,we designed a TWK-means(Trust Weight K-means)clustering algorithm based on resource attributes,which improves the reliability and efficiency of the service by quickly and accurately clustering the resource required for the task.Finally,a task offloading model based on TWK-means clustering is constructed to enhance the efficiency of the network service system.The experimental results show that the clustering efficiency of the proposed TWKmeans algorithm is improved by 20%,and the devices required by the task are obtained quickly.At the same time,for the final offloading target,the overall delay and energy loss based on the TWK-means clustering offloading model will be reduced by 13%.(3)Aiming at the optimization of task offloading in edge computing environment,a trust-aware intelligent offloading algorithm based on deep reinforcement learning is proposed.The edge computing paradigm solves the limitation of the computing power of edge terminal devices for computingintensive and delay-sensitive tasks through computing offloading technology.However,much existing works focus on optimizing policies with the goal of reducing service latency and energy consumption,and most of the works ignore service reliability,which makes nodes vulnerable to attacks.First,we propose a trusted-aware intelligent offloading architecture based on DRL in an edge computing environment.The architecture achieves vertical collaboration between devices and servers,horizontal collaboration between devices and devices,which solves the contradiction between the limited computing power of heterogeneous devices and the diversity of task demands in the case of service reliability,which realizes availability and scalability of the network service system.Then,a multi-feedback trust fusion model based on temporal difference is designed,which can effectively overcome the limitations of artificial weighting and effectively resist the malicious attacks of nodes in the network service environment.Finally,in view of the dynamics,stochastic and complexity of the network,the task offloading problem is modeled as a Markov decision process,and a trust-aware intelligent offloading algorithm based on DRL is proposed,which realizes the flexible offloading of tasks.From the experimental results,compared with the existing schemes,the proposed multi-feedback trust model based on Temporal Difference improves the system response by 2-4 times and the malicious detection rate by 60%.Compared with the mainstream offloading policy,the proposed trust-aware offloading algorithm reduces the delay and energy loss by 6%,which not only ensures the safe operation of the system but also promotes the service efficiency. |
PDF Full Text Request