Article Preview
TopIntroduction
Various studies have established the fact that the number of lives lost in motor vehicle crashes world-wide every year is by far the highest among all the categories of accidental deaths (Wikipedia, n.d.). With the expected increase in the vehicle and human populations as well as economic activities, roads will likely get busier. Thus, there is an urgent need to enhance road safety and reduce traffic congestion. Recently, with the advancement in technology more and more vehicles are being equipped with GPS and Wi-Fi devices that enable vehicle to vehicle (V2V) communication, forming a vehicular ad-hoc network (VANET). Peer vehicles in VANET can communicate with each other regarding up to date information about road and traffic conditions, so as to avoid car accidents and effectively route traffic through dense urban areas. VANET is thus envisioned to be one of the most important applications of mobile ad-hoc networks in the future.
Network-On-Wheels (NOW) project (http://www.car-to-car.org/) among others, represent some of the ongoing efforts in the general domain of vehicular networks. Some car manufacturers have already started to fit devices that will help achieve the goals mentioned above. For example, GM has rolled out V2V communication in its Cadillac STS Sedans. GM’s proprietary algorithm called “threat assessment algorithm” keeps track of the relative position, speed and course of other cars (also equipped with V2V technology) in a quarter-mile radius and issues a warning to the driver when a crash is imminent. Similar prototypes by other car manufacturers are currently in the testing phase, scheduled to hit the markets over the coming years (Nadeem et al., 2004; Xu et al., 2004; Elbatt et al., 2006; Rahman & Hengartner, 2007). These systems focus mainly on ensuring a reliable delivery of messages among peers. As a result, less focus has been placed on evaluating the quality of information that is sent by peers, in order to cope with reports from malicious peers which may compromise the network. For example, consider a peer that reports the roads on his path as congested with the hope that other peers would avoid using these roads, thus clearing the path. Therefore one important issue among others that may arise in VANETs is the notion of trust among different peers.
The goal of incorporating trust is to allow each peer in a VANET to detect dishonest peers as well as malicious data sent by these dishonest peers, and to give incentives for these peers to behave honestly and discourage self-interested behavior. Given the critical nature of the applications in the context of VANETs, it is crucial to associate trust with peers and the data that they spread. However, due to the important and possibly unique characteristics of VANET environments, effectively modeling trust of peers becomes very challenging.
In this article, which is an extended version of Zhang (2011), we first discuss the challenges for trust management caused by the large, decentralized, open, sparse and highly dynamic nature of VANET environments, and identify some key desired properties that trust management should incorporate, including decentralized trust establishment, being capable of coping with sparsity, being event/task and location/time specific, scalable, robust and sensitive to privacy concerns, and integrated confidence measure. For each property, we also extensively discuss some trust models proposed in other domains (such as multi-agent systems, peer-to-peer systems, collaborative intrusion detection networks, etc.) that may provide useful solutions. We then survey and evaluate the existing trust models in VANETs based on the desired properties. None of them has achieved all the properties. We finally suggest some important future research directions towards effective trust management in VANETs.