Internetware is an abstract of the distributed software system in the open and dynamic Internet which is generally integrated by large numbers of autonomous and heterogeneous software services of various functions. The reliability of the service which acts as a component in Internetware inevitably has an effect on the collaboration in the system and users' satisfaction on the whole system. So picking the right service from the open environment is a critical problem for developers to build high confidence application systems. However, traditional test measures cannot work when facing with dynamically evolutive characters and separated interests between developers of systems and third-party services in the Internetware. Trust and reputation mechanism is a complementary approach which relies on analyzing collected recommendations from past users to evaluate the software. In this paper, we propose a personalized trust-based approach to do service selection automatically for users. It mainly deals with two problems: 1) trustworthiness: as lacking of a global monitor to record runtime performances of services, we design a recommendation collection method based on the user's social network; 2) relevance: since the formation of ratings has an impact on the recommendation, we establish a filtering process to weight information according to recommenders' expectation and capability.
展开▼
