High Speed Data Routing in Vehicular Sensor Networks Harry Gao, Seth Utecht Department of Computer Science, College of William and Mary Gregory Patrick, George Hsieh Department of Computer Science, Norfolk State University Fengyuan Xu, Haodong Wang, Qun Li Department of Computer Science, College of William and Mary Abstract—In this paper, we show through a simple secure symmetric key based protocol design and experiments the feasibility of secure data collection in a vehicular sensor networks. This protocol exhibits high speed data rout- ing for sensor data collection through vehicles. The large communictaion and storage capacities of a vehicle and its mobility facilitates this high speed routing scheme compared with routing through hop-by-hop communication among sensors. We demonstrate that the protocol works in a realistic setting by collecting the real trace data through real implementation. I. I NTRODUCTION Vehicular network has attracted people’s attention in recent years with the vision that it can provide crucial in- formation, such as traffic conditions, to interested parties. The vehicular network architecture is mainly composed of inter-vehicular communications, and communication between vehicles and the roadside sensors. Although the picture is very exciting, we do not expect the technology to be mature in the next couple of years for very practical deployment. This is due to the hurdles along the way: the standardization of the network communications, the effort associated with the deployment of the roadside sensors, and the maturity of the hardware. These processes can be both expensive and time consuming. We argue in this paper that a simple architecture based on sensors (e.g., Berkeley Motes) can fulfill many impor- tant functions envisioned in vehicular networks. Sensors are deployed along the roadside to collect environmental data. For example, the sensors can gather data on highway conditions (e.g. potholes, cracks on the road, ice on the road and blind spots ahead). They can also monitor the environment for scientific purposes, such as monitoring pollution or pollen count. Moreover, they can be used as a temporary storage space for data. For instance, if a vehicle notices a collision ahead, it can send a message to the roadside sensor so that the vehicles behind may know the information when they are within the transmission range of the sensor. This architecture becomes more powerful when the vehicles are harnessed to carry data stored or collected in a sensor to the more sophisticated servers deployed at weigh stations, toll gates, or rest areas. This information can then be processed, analyzed and broadcasted, and can be made available to the general public through services such as Google Map. A small network following this architecture is inexpensive and easy to deploy, because the price of motes continues to drop. A vehicle simply needs to be enhanced with the capability of communicating with sensors. In this network architecture, it is crucial to provide security support – only authorized vehicles can feed data into sensors and to obtain data from sensors. To block unauthorized and malicious vehicles, data collected by sensors must be encrypted. However, merely encrypting the data cannot prevent a malicious car to obtain the scrambled data. Although the encrypted data is of little use to the malicious vehicle, it is a serious problem when sensors expect vehicles to harvest all the data and carry them to a central station; a malicious vehicle can simply trap the data and leave a hole in the designated data repository. Therefore, authenticating a passing vehicle before transferring any data is indispensable. A straightforward solution is to use a public-key based scheme, since some (e.g., the ECC) of the schemes can be implemented efficiently on sensor platforms. Taking a closer look at the problem in a real experimental study, we found that authentication takes about one to two seconds in many cases, which is non-negligible for a car traveling at high speed. A car may rush out of a sensor’s transmission range after the authentication is conducted. In this paper, we show our security solution to the vehicular sensor networks and give experimental results on a realistic deployment. We show through a simple secure protocol design the feasibility of secure data collection in a vehicular sensor networks. We deployed sensors along the roadside to test the performance of the communication between the roadside sensors and the sensors in a moving vehicle. We demon- strate the protocol works in a realistic setting by collecting the real trace data through real implementation. We hope this research shows valuable experience in deploying security support for this type of networks. II. RELATED WORK There already exist many proposals for some of the challenging aspects, such as the session layer protocol Manuscript received May 10, 2009; accepted Nov. 17, 2009. JOURNAL OF COMMUNICATIONS, VOL. 5, NO. 3, MARCH 2010 181 © 2010 ACADEMY PUBLISHER doi:10.4304/jcm.5.3.181-188