Research On Trust Evaluation Model Based On Graph Theory For Edge Computing Environment

Edge computing uses many edge devices to provide users with real-time computing and storage functions near the ground.It can meet the needs of users with low delay and fast response,and reduce the pressure of cloud computing center and network bandwidth.However,the requirements of collaboration between edge devices are different,and edge computing has the characteristics of open autonomy and multi-source heterogeneity,which leads to the lack of necessary trust between devices.It is difficult to resist the internal malicious attacks such as service manipulation and data tampering.The trust mechanism can effectively resist the internal attacks of the network and ensure the reliable services provided by the devices.However,the massive trust information in the edge layer forms a huge trust network,and the edge devices with limited resources are difficult to undertake complex storage and query tasks.Therefore,how to build a lightweight trust model in the edge computing environment has become a research hotspot.This thesis puts forward solutions to the above problems,and the main research work is as follow:(1)In order to reduce the resource overhead of the edge computing trust model and solve the problem of trust path redundancy,an edge computing trust evaluation model based on DFS algorithm was proposed.Firstly,the complex and huge trust relationship between devices was abstracted into a directed weighted graph,which was defined and explained.And an adaptive aggregation method based on information entropy theory was used to aggregate the trust value;Secondly,by adding multiple constraints such as trust threshold,path length limit and sliding window,the nodes and trust edges that do not meet the requirements were filtered.Finally,the improved depth first search algorithm was used to optimize the redundant trust path to avoid the loop and node bypass problems.The experimental results showed that the algorithm could not only guarantee the validity of the trust model,but also reduce the resource overhead of the edge devices in the process of trust evaluation.(2)In order to improve the accuracy of trust evaluation,ensure the load balance within the domain,and solve the problems of single evaluation attributes and highly overlapping trust paths,a multi-attribute evaluation model based on KSP algorithm was proposed.Firstly,from the view of multi-attribute of service,the attribute of trust evaluation was refined,and the weight was allocated by AHP to meet the special needs of individual users.Secondly,based on KSP(K Shortest Paths)penalty factor algorithm,a trust path optimization algorithm RKSP(KSP based on Repeatability)was proposed.It could filter highly overlapped paths in the process of trust path formation,make the search paths have certain differences.And by using the characteristics of heuristic search of A* algorithm,the number of visited nodes was reduced.Experimental results showed that the model could reduce the resource overhead of edge devices,and make the results of model recommendation more in line with the needs of users.