SUMMARY
The accuracy of feedback presentation, the sensitivity, and the robustness of reputation evaluation are critical issues to be addressed in a reputation system under the grid environment. This paper proposes a QoS‐based fine‐grained grid reputation system, where economic elements are considered to make the reputation system more sensitive in the commercial grid environments. A novel fine‐grained feedback presentation model based on aggregation of objective QoS attributes and subjective opinions of evaluators is proposed to enable a semi‐automatic, personalized and accurate feedback presentation. Through the introduction of a punishment factor and the adaptive reference of a previous trust value, the proposed reputation system effectively improves the sensitivity of the reputation evaluation. Moreover, the weighted combination of interorganizational trust, direct trust and recommended trust makes the reputation system more robust against collusion attacks. Simulation results show that the proposed reputation system can practically predict feedback and effectively resist malicious attacks such as fake transaction attacks and badmouthing attacks. This provides a clear advantage in the applications of grid service selection. Copyright © 2011 John Wiley & Sons, Ltd.