Reputation assessment mechanism for carpooling applications based on clustering user travel preferences

Abstract

One way to ensure sustainable and environmental friendly mobility is the use of less vehicles for carrying more passengers, and carpooling is a means to achieve this goal. One major concern in carpooling services is related to trust, as carpooling users need to either share their vehicles, if they act as drivers, or travel with strangers if they act as passengers. One way to tackle trust concerns is the utilization of user reputation assessment mechanisms, whose objective is to provide ranking of users with respect to their behavior, based on feedback provided by other users. This paper presents a newly introduced reputation assessment mechanism for carpooling applications, which, in addition to feedback provided by other users, takes into account user travel preferences. Preliminary experimental results have shown that the proposed mechanism is robust against attacks by malicious users, on their attempt to jeopardize the trustworthiness of the system, as it preserves the real reputation scores of the users, when the penetration rate of malicious users increases. Keywords: Carpooling, Reputation systems, Clustering, k-means