QoS-Based Reputation Feedback Fusion under Unknown Correlation

Abstract

Abstract. Due to the flood of web services that offer similar functionalities, service consumers are left with a challenging selection decision. A popular approach to assist them with the service selection task is based on the reputation of web services. However, the propagation of reputation feedback in an open and distributed system of web services yield correlated reputation estimates. The existing web service reputation literature still lacks a system that handles the aggregation of reputation feedback under unknown correlation. To fill this gap, we employ two data fusion algorithms, the covariance intersection and ellipsoidal intersection, to aggregate QoS-based reputation feedback. Our experimental results endorse the advantageous capability and scalability of the proposed methods in aggregating reputation estimates, and show an enhanced performance when compared with the Kalman filter method.