| Along with the development of the Internet, the characteristics of the multiplicity of identity, the randomness of network transactions and the dynamic and dispersion of network data for the network entities proposes the new need and challenges for the network security. How to ensure the security, effective and sustainable development of internet transactions becomes the important issues of concern increasingly. Meanwhile it continually puts forward the new research directions for the trust management model which described the different network scenarios. Because of the traditional access control and other techniques are unable to meet the development needs of the existed network gradually, many researchers are continuing to set new management models to solve the network problems. How to calculate the recommendation trust which is one considered aspect in trust model quickly, reasonably, objectively and comprehensively provides an important basis and security guarantees for selecting a recommended chain based more accurate information from the request entity to the provide entity, and it also has become one of the most prominent research focus in the trust management model studies. The study for trust management model used in existed network, especially the study for the recommendation trust computing and recommended chain selection, has great significance to ensure the security of the interest of the entities in the network.Based on existed researches for the trust management model and after considering the recommendation accuracy and its positive or negative impact for network transactions, a new trust management model through introducing the concept of confidence intervals for improving recommendation accuracy is given in this paper. The concepts of opportune deal time and overtime deal time given out first in this model give a more reasonable judgment for determining whether the transaction is completed and solving the judgment problem that the time required of some transactions beyond the system set time. If a transaction can be finished during the opportune deal time and overtime deal time, it still can be regarded as a successful completed transaction. In the process of recommendation, the model considers the different emphasis for the indicators of the same services for the different entities. According to the practical requirements of the entity, the model will recount the recommendation which is provided by the other entities before the transaction occurs and calculate the recommended reference trust combined with the emphasis. This method can take full account of the subjective differences between different entities. In the feedback evaluation for the recommendation entities, this model introduces the concept of confidence interval to calculate the allowed range of the number of unsuccessful recommended. It allowed the recommendation trust value fluctuate within the certain range to avoid the feedback error given by recommendation entities because of their subjective difference. In addition, when the number of unsuccessful recommended is accumulated to a given value, the punishment mechanism will be enabled. This mechanism will not modify the trust of entities only because of one unsuccessful recommendation, and encourage entities to participate recommendation more actively. The experimental results show that the model can consider all kinds of transaction data reasonably and express the difference of same recommendation trust for different entitiesBased on the exited mechanisms for measuring the recommendation, a directed graph algorithm with minimal risk which combines the characteristics of existing trust networks and the idea of directed graph is put forward for selecting the recommended chain. Without the desired rule, the algorithm dilutes the assumptions that the recommended chain must be independent with each other and expresses the recommendation relationship with the directed graph. It allows the existence of cross-entity and shows the real recommendation relationships of all entities involved in recommended chain. Hence, in the data and relationships of large complex networks, it can better handle the issue that some entities recommended others more than once in a recommended relationship. In the process of selecting recommended chain, the algorithm puts the recommended relationships to a directed graph, and finds the best recommended path through calculating the risk value of the recommended chain using the two entities’similarity, local trust, global trust, transaction history, etc. The algorithm seeks the recommended chain with minimum risk from request entity to target entity through the information in the directed graph constituted by the recommendation relationship and the risk of every side. The request entity can select the target entity based on the recommended chain got from the algorithm or re-run the algorithm to select the recommended chain by adjusting the risk of every side for several times according their own risk tolerances and other factors. In addition, when the number of entities formed a directed graph is less than a given threshold, the virtual recommendation will be introduced for adding the recommended entities in order to expand the use of the algorithm. The risk of the virtual recommendation chain computed from the two entities’direct trust, local trust, global trust, transaction history, etc. will be joined into the algorithm to select the recommendation path. This algorithm objectively considers most of the recommended information of entities included in the recommended chain. |
PDF Full Text Request