Determining network availability on the move

Users today have come to expect constant connectivity through their mobile devices even when they are on the move, as most urban areas are heavily blanketed with Access Points and GSM Base Stations. Network selection poses a problem however as it is driven primarily by physical layer information, such as Received Signal Strength, which is a poor indicator of actual network performance. Thus there is a need for a more sophisticated method of discovering network resources that goes beyond radio interface characteristics. A way to address this is to use information not only from within the node, but also externally, for instance by enabling nodes to exchange information in a peer to peer manner. One of the biggest challenges of adopting this approach stems from the fact that nodes participating in such systems are not equally trustworthy. Therefore it is necessary to develop mechanisms that make the decision making process robust against dishonest information. In this work we evaluate data fusion techniques in the context of reconciling conflicting information due to the presence of dishonest or malfunctioning nodes. We show that it is possible to adapt classical data fusion techniques to develop a more robust decision-making mechanism, and moreover that Dempster-Shafer Theory is the optimal choice.