Reputation systems and secure communication in vehicular networks

A thorough review of the state of the art will reveal that most VANET applications rely on Public Key Infrastructure (PKI), which uses user certificates managed by a Certification Authority (CA) to handle security. By doing so, they constrain the ad-hoc nature of the VANET imposing a frequent connection to the CA to retrieve the Certificate Revocation List (CRL) and requiring some degree of roadside infrastructure to achieve that connection. Other solutions propose the usage of group signatures where users organize in groups and elect a group manager. The group manager will need to ensure that group members do not misbehave, i.e., do not spread false information, and if they do punish them, evict them from the group and report them to the CA; thus suffering from the same CRL retrieval problem. In this thesis we present a fourfold contribution to improve security in VANETs. First and foremost, Chains of Trust describes a reputation system where users disseminate Points of Interest (POIs) information over the network while their privacy remains protected. It uses asymmetric cryptography and users are responsible for the generation of their own pair of public and private keys. There is no central entity which stores the information users input into the system; instead, that information is kept distributed among the vehicles that make up the network. On top of that, this system requires no roadside infrastructure. Precisely, our main objective with Chains of Trust was to show that just by relying on people¿s driving habits and the sporadic nature of their encounters with other drivers a successful reputation system could be built. The second contribution of this thesis is the application simulator poiSim. Many¿s the time a new VANET application is presented and its authors back their findings using simulation results from renowned networks simulators like ns-2. The major issue with network simulators is that they were not designed with that purpose in mind and handling simulations with hundreds of nodes requires a massive processing power. As a result, authors run small simulations (between 50 and 100 nodes) with vehicles that move randomly in a squared area instead of using real maps, which rend unrealistic results. We show that by building tailored application simulators we can obtain more realistic results. The application simulator poiSim processes a realistic mobility trace produced by a Multi-agent Microscopic Traffic Simulator developed at ETH Zurich, which accurately describes the mobility patterns of 259,977 vehicles over regional maps of Switzerland for 24 hours. This simulation runs on a desktop PC and lasts approximately 120 minutes. In our third contribution we took Chains of Trust one step further in the protection of user privacy to develop Anonymous Chains of Trust. In this system users can temporarily exchange their identity with other users they trust, thus making it impossible for an attacker to know in all certainty who input a particular piece of information into the system. To the best of our knowledge, this is the first time this technique has been used in a reputation system. Finally, in our last contribution we explore a different form of communication for VANETs. The vast majority of VANET applications rely on the IEEE 802.11p/Wireless Access in Vehicular Environments (WAVE) standard or some other form of radio communication. This poses a security risk if we consider how vulnerable radio transmission is to intentional jamming and natural interferences: an attacker could easily block all radio communication in a certain area if his transmitter is powerful enough. Visual Light Communication (VLC), on the other hand, is resilient to jamming over a wide area because it relies on visible light to transmit information and ,unlike WAVE, it has no scalability problems. In this thesis we show that VLC is a secure and valuable form of communication in VANETs