Securing Cluster Formation and Cluster Head Elections in Wireless Sensor Networks

In wireless sensor networks, clustering plays a very important role for energy savings at each node because it reduces the number of transmissions through TDMA based communication. For secure clustering, it is very crucial to find compromised nodes and remove them during the initial cluster formation process. If some nodes are compromised and survive from the exclusion process of normal nodes, they can make some nodes have a different membership view in the same cluster and consequently separate a cluster into multiple clusters. To resolve these problems, we propose a robust scheme against such attacks in this paper. First, our scheme generates large sized clusters to improve the quality of clusters. Second, our scheme exploits the verification of two hop distant nodes to maintain the quality of the large sized clusters and avoids the separation of the clusters. In addition, our scheme prefers broadcast transmissions to reduce the energy consumption of nodes. We prove that our scheme generates fewer clusters and is more secure and energy-efficient than its rival scheme through security analysis and simulation results. With regard to CH election, we also propose a scheme which securely elects CHs by recognizing the compromised nodes and depriving them of their CH candidacy. To this aim, each node in a cluster calculates reputation values of other CH candidates according to their behavior and distributes them through a broadcast. Then each node extracts substantial reputation values of CH candidates using the distributed reputation values. Next, each node evaluates the substantial reputation values of other CH candidates and excludes some disreputable nodes from CH candidates. The scheme greatly improves non-manipulability and agreement property of CH election results in comparison with other rival schemes. Moreover, the scheme guarantees higher non-manipulability and agreement property than other rival schemes, even in a loss-prone environment.