Comparing Local, Collective, and Global Trust Models

Abstract—Today, trust modelling is a serious issue on the social web. Social web allows anonymous users to exchange information without even knowing each other beforehand. The aim of a trust model is to rerank acquired information according to their reliability, and to the trustworthiness of their authors. During the last decade, trust models were proposed to assist the user to state his opinion on the acquired information, and on their sources. We identify three paradigms for trust modelling: the first relies on evaluating previous interactions with the source (individual trust), the second relies on the word of mouth paradigm where the user relies on the knowledge of his friends and their friends (collective trust), and the third relies on the reputation of the source (global trust). In this paper, we propose and compare three trust models, each of which represent one of the precedent paradigms. All three models make use of subjective logic (SL). SL is an extension of probabilistic logic that deals with the cases of lack of evidence. It supplies framework for modelling trust on the web. The comparison includes three axes: the precision, the complexity and the robustness to malicious attacks. We show that each of the three models has a weak point in one of the three axes