Fairness-guaranteed reputation propagation in Web service composition

In service-oriented environment, it is difficult to evaluate component services because of the opaque characteristic of composite services, the complex invocation structures and the subjective reputation rating of service consumers. To address these issues, this paper proposes a reputation propagation algorithm for service composition, in which the subjective ratings can be fairly propagated to each component service. The algorithm first models service composition as the Beta-mixture, and learns the reputation and responsibility of each component by the EM algorithm. Then, based on the characteristics of Shapley values in cooperative gaming theory, the algorithm computes the contribution of each component to its composition, ensuring that no component would obtain extra rewards or punishments. Finally, theoretical analysis and experimental results demonstrate the fairness of the algorithm to hieratically propagate the consumer's rating to each component service.