A distributed sinkhole detection method using cluster analysis

In recent years, the popularity of wireless devices has grown dramatically. Mobile ad hoc networks (MANETs) are considered to be vitally important for wireless communication. However, the dynamic nature of MANETs makes their routing protocols vulnerable to attacks. In this paper, we focused on sinkhole attacks, representative of routing-disruption attacks. First, we analyzed sinkhole attacks under a dynamic source routing protocol. Unlike wired networks, where routing-disruption attacks can be prevented through authentication mechanisms, application of authentication techniques in wireless networks is nontrivial—the lack of centralization hampers authentication support management. We then used a data mining technique called cluster analysis to develop a sinkhole attack detection method in wireless networks. The cluster analysis to route request packets does not require any centralized infrastructure. Robust features for detecting sinkhole attacks in a distributed manner are suggested. Simulation shows excellent classification performance and successful distributed detections.