A model implementing certified reputation and its application to TripAdvisor

Many real-life reputation models suffer from classical drawbacks making the systems where they are used vulnerable to users' misbehavior. TripAdvisor is a good example of this problem. Indeed, despite its popularity, the weakness of its reputation model is resulting in loss of credibility and growth of legal disputes. In this paper, we propose a reputation model abstractly considering service providers, users and feedbacks, and implementing the theoretical notion of certified reputation to concretely define a strategy to normalize feedback scores towards reliable values. We apply the model to the case of TripAdvisor, by proposing a solution to improve its dependability not increasing invasiveness nor reducing usability of the system. Moreover, it fully guarantees backward compatibility. In the context of project activities, we are in progress to fully implement the system and validate it on real-life data.