A convex optimization approach for clone detection in wireless sensor networks

When deployed unattended in hostile environments, static and mobile Wireless Sensor Networks (WSNs) are vulnerable to node capture and cloning attacks, where an adversary physically compromises network nodes and extracts all information known to them, including the assigned cryptographic material and the internal states of network protocols. The obtained knowledge is used to disrupt the network by deploying and controlling copies of captured nodes (clones). Recently, a variety of novel clone detection methods have been developed, using concepts such as birthday paradox, sequential detection or random encounters in mobile environments. At present there is no framework to evaluate an individual detection method based on the WSN performance under attack or to compare and choose a method appropriate for a given application. In this paper, we develop an optimization framework for choosing the parameters of a detection method so that the cost of clone detection is minimized. We show that every detection method can be characterized in terms of four costs, namely, the impact of leaving undetected cloned nodes in the network, the cost of revoking nodes falsely identified as compromised, and the costs of communication and storage. A convex combination of these costs defines the cost of clone detection, which is then minimized with respect to the parameters of the detection method. Using this framework, we analyze existing clone detection algorithms and provide efficient methods for obtaining optimal detection parameters.