Detecting node replication attacks in mobile sensor networks: theory and approaches

A wireless sensor network composed of a number of sensor nodes is often deployed in unattended and harsh environments to perform various monitoring tasks. Due to cost concerns, usually, sensor nodes are not made tamper-resistant, and a captured node may be easily compromised by an adversary. With the revealed secret credentials, the adversary can create many duplicate nodes that are seemingly legitimate, and deploy them into the network to cripple the monitoring applications. Defending against node replication attacks has become an important research topic in sensor network security, but so far, not many solutions have been proposed, most of which adopt a stationary network model where sensor nodes are fixed and immobile. In this work, we address the problem of detecting node replication attacks in a mobile sensor network, where each sensor node freely and randomly roams in the sensing region all the time, and one node meets with another in an occasional and unpredictable manner. For replication attacks where the replicas do not conspire, we employ very lightweight token-based authentication as a detection approach. In case the replicas conspire by communicating with each other in an efficient manner, we harness the random encounters between physical nodes and propose a detection method based on statistics. Compared with existent solutions, our detections have the nice feature that sensor nodes are freed from the fragile assumption that they can correctly obtain their geographic positions, and that even loose time synchronization may be unnecessary. Copyright © 2011 John Wiley & Sons, Ltd.