This paper presents a novel approach for detecting faulty sensors as well as identifying the type of faults with a high degree of confidence using reputation concepts and introducing special nodes as beacon nodes to the network and with very low false alarm rate. Our algorithm is purely localized and thus scales well to large sensor networks as well as being efficient with respect to the amount of generated traffic and resource usage. The computational overhead for a sensor is low, since only simple operations are involved in each sensor node and main tasks are done in beacon nodes which have a lot of resources and number of them is low. The complexity of the algorithm is very low and the probability of correct diagnosis is very high even in the existence of large fault sets. Simulation results show the algorithm can clearly identify the faulty sensors with high degree of trust and low communications for sensors.