In civil aviation applications, Satellite failures yield unacceptable positioning errors when using the Global Positioning System (GPS). To ensure the user security, the navigation system has to fulfill stringent performance requirements. Thus, detecting and excluding the faulty GPSmeasurements is necessary prior to estimating the mobile location. Classical fault detection algorithms based on Kalman filters (KF) are sensitive to the choice of an appropirate motion model for the mobile. To overcome this difficulty, we propose in this paper a new fault detection algorithm wherein the KF are replaced by multiple model algorithms. In this way, both the false alarm rate and the positioning mean square error are shown to be decreased.