A Perceptron Based Classifier for Detecting Malicious Route Floods in Wireless Mesh Networks

Wireless mesh networks (WMN) are evolving as a new paradigm for broadband Internet, in which a group of static mesh routers employ multihop forwarding to provide wireless Internet connectivity. All routing protocols in WMNs naively assume nodes to be non- malicious. But, the plug-in-and-play architecture of WMNs paves way for malicious users who could exploit some loopholes of the underlying routing protocol. A malicious node can inundate the network by conducting frequent route discovery which severely reduces the network throughput. In this paper, we investigate the detection of route floods by incorporating a machine learning technique. We use a perceptron training model as a tool for intrusion detection. We train the perceptron model by feeding various network statistics and then use it as a classifier. We illustrate using an experimental wireless network (ns-2) that the proposed scheme can accurately detect route misbehaviors with a very low false positive rate.