From Eigentrust to a Trust-Measuring Algorithm in the Max-Plus Algebra

Juan Afanador, Nir Oren, Murilo S. Baptista, Maria Araujo

Abstract: Eigentrust is a simple and popular method for trust computation, which uses both direct and indirect information about individual performance to provide a global trust rating. This final trust value is based on eigenvectors computed through the Power Method. However, under certain network topologies, the Power Method cannot be used to identify appropriate eigenvectors. After characterising these cases, we overcome Eigentrust’s limitations by extending the algorithm’s core ideas into the Max-Plus Algebra. An empirical evaluation of our new approach demonstrates its superiority to Eigentrust.