SelfTrust： leveraging self-assessment for trust inference in Internetware

Abstract Internetware is envisioned as a new software paradigm for software development in platforms such as the Internet. The reliability of the developed software becomes a key challenge due to the open, dynamic and uncertain nature of such environment. To make the development more reliable, it is necessary to evaluate the trustworthiness of the resource providers or potential working partners. To this end, we propose a novel trust inference approach to evaluating the trustworthiness of potential partners to guide the software development in Internetware. The main insight of our approach is to employ the self-assessment information in order to improve the trust inference accuracy. Especially, we first extend the balance theory and the status theory from social science to incorporate self-assessment, and then propose a machine learning framework to extract several features from the extended theories and infer trustworthiness scores based on these features. Experimental results on a real software developer network show that the self-assessment information truly helps to improve the accuracy of trust inference, and the proposed SelfTrust model is more accurate than other state-of-the-art methods.