A cooperative watchdog model based on Dempster–Shafer for detecting misbehaving vehicles Omar Abdel Wahab a , Hadi Otrok b , Azzam Mourad a,⇑ a Department of Computer Science and Mathematics, Lebanese American University, Beirut, Lebanon b Department of Electrical & Computer Engineering, Khalifa University of Science, Technology & Research, Abu Dhabi, United Arab Emirates article info Article history: Received 10 April 2013 Received in revised form 17 October 2013 Accepted 22 December 2013 Available online 7 January 2014 Keywords: Vehicular Ad hoc Network (VANET) Dempster–Shafer Cooperative detection Reputation Passive malicious nodes abstract In this paper, we address the problem of detecting misbehaving vehicles in Vehicular Ad Hoc Network (VANET) using Quality of Service Optimized Link State Routing (QoS-OLSR) protocol. According to this protocol, vehicles might misbehave either during the clusters’ formation by claiming bogus information or after clusters are formed. A vehicle is considered as selfish or misbehaving once it over-speeds the maximum speed limit or under-speeds the minimum speed limit where such a behavior will lead to a disconnected network. As a solution, we propose a two-phase model that is able to motivate nodes to behave cooperatively during clusters’ formation and detect misbehaving nodes after clusters are formed. Incentives are given in the form of reputation and linked to network’s services to motivate vehicles to behave cooperatively during the first phase. Misbehaving vehicles can still benefit from network’s ser- vices by behaving normally during the clusters’ formation and misbehave after clusters are formed. To detect misbehaving vehicles, cooperative watchdog model based on Dempster–Shafer is modeled where evidences are aggregated and cooperative decision is made. Simulation results show that the proposed detection model is able to increase the probability of detection, decrease the false negatives, and reduce the percentage of selfish nodes in the vehicular network, while maintaining the Quality of Service and stability. Ó 2013 Elsevier B.V. All rights reserved. 1. Introduction Vehicular Ad Hoc Network (VANET) [16,20,19,9] is a new kind of ad hoc networks that is characterized by its highly mobile topol- ogy. Like Mobile Ad hoc Network (MANET), VANET encounters the problem of selfish nodes that may hinder the implementation of any protocol dedicated to it. However, dealing with these nodes in VANET is more challenging due to the increased ambiguity in the detection caused by the high mobility of vehicles. The Quality of Service Optimized Link State Routing (QoS-OLSR) protocol [10] is a proactive routing protocol modeled to cope with mobile ad hoc networks. It is based on electing a set of optimal cluster-heads and dividing the network into clusters. These heads are then responsible for selecting a set of designated nodes charged of transmitting the network topology information and forwarding the traffic flows. Such nodes are called MultiPoint Relay (MPR) nodes. This protocol is an enhanced version of QOLSR [1] that pro- longs the network lifetime by considering the energy of nodes while calculating the QoS function since the nodes, in MANET, have limited energy resources. However, the energy parameter has a minimal importance in VANET due the long battery lifetime of vehicles. In order to extend such a protocol to VANET, velocity and residual distance parameters must be added to the QoS func- tion instead of the residual energy to improve the network stability. According to this protocol, vehicles might misbehave either during the clusters’ formation by claiming bogus information or after clusters are formed. A vehicle is considered as selfish or mis- behaving once it over-speeds the maximum road limit or under- speeds the minimum road limit. Such a behavior is considered as a passive malicious since vehicles do not aim to attack or impede the network functioning, but rather they tend to optimize their own gain neglecting the welfare of others [11]. They entail hence negative implications on the whole network such as the (1) in- crease in the percentage of MPRs, (2) decrease in the network sta- bility, (3) increase in the clusters disconnections, and (3) increase in the average path length. To address the above problems, we propose a two-phase model that (1) motivates vehicles to behave normally during clusters’ formation and (2) detects misbehaving vehicles after clusters’ formation. In phase one, incentives are given in the form of reputation where networks’ services are offered based on vehicle’s 0140-3664/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.comcom.2013.12.005 ⇑ Corresponding author. Tel.: +961 (1) 786456x1200; fax: +961 (1) 867 098. E-mail addresses: omar.abdelwahab@lau.edu.lb (O.A. Wahab), Hadi.Otrok@ kustar.ac.ae (H. Otrok), azzam.mourad@lau.edu.lb (A. Mourad). Computer Communications 41 (2014) 43–54 Contents lists available at ScienceDirect Computer Communications journal homepage: www.elsevier.com/locate/comcom