Research On Group-based Trust Model In Mobile P2P Network

With the continuous development of network technology,P2 P networks have been widely used due to their anonymity and openness,such as file sharing and instant communication.At the same time,these characteristics also bring new challenges to P2 P networks.Malicious peers will spread illegal messages at will,which will have a negative impact on the availability and user experience of the network,In addition,the dynamics and distribution of P2 P networks also make it difficult for peers to conduct long-term and stable transactions.The current trusted model does not solve the problems of unreliable transactions and dynamic network topology in P2 P networks.The existing research results still have insufficient results.For example,in the existing trust model algorithms based on P2 P groups,peers are grouped only based on their similarity,and the grouping standard is single,which makes the system lack robustness.And traditional trust models generally acquire transaction peers based on recommended algorithms,but this kind of trust models have large computational overhead and high complexity.In view of the shortcomings listed above,this paper proposes a group-based trust models in mobile P2 P networks.The paper's innovations and work are summarized as follows:First,a new type of dual dynamic group trust model named Dual Trust is proposed.First,the peers are divided into several large groups through similarity calculation,and then into several groups based on the relative distance in the large group.Each peer is given a different role,and three new types of trust measurement methods are proposed.The problem of super peer selection is improved,and a new solution strategy is proposed for the dynamic joining and leaving of peers.Experiments prove that the new model has a higher successful transaction rate and lower communication overhead when processing large-scale data.Secondly,a grouping algorithm based on DBSCAN is proposed for P2 P network.Based on the idea of parallel computing,the peers are partitioned firstly,and the similarity between peers is used as the basis for partitioning.The k-distance method and kernel function are introduced into the parameter selection in order to avoid the impact of the manual intervention grouping results,and then group merging is carried out after local clustering.Since peer grouping belongs to unsupervised classification,in the experimental part,running time,contour coefficient and DB index are introduced as evaluation criteria.Experiments show that the grouping algorithm has better classification effect.Finally,a new trust algorithm Ga-Bidding is proposed to solve the problem how to select trusted transaction peers in the P2 P group.This algorithm refers to the bidding process of human society,firsty,a group of guarantee peers are applied for each service peer to guarantee the reputation of the service.Secondly,based on the entropy TOPSIS selection method,a set of ideal transaction sequences is selected.Then,the transaction influence function is used to assign weights to the guarantee trust degree of the guarantee peer sequence and the global trust degree of the service peer,so as to calculate the comprehensive trust degree of each service peer.Finally,based on a specific confidence level,the service peer is verified by feedback,and the service peer and its guarantee peer sequence are updated for reputation.Experiments show that Ga-Bidding has a good effect in improving the successful transaction rate and defending against complex attacks.