Resilient Cluster Leader Election for Wireless Sensor Networks

Abstract

Abstract-Sensor nodes are often organized into clusters for efficiency and scalability purposes. Every sensor cluster is managed by a cluster leader during the network operation such as routing and data aggregation. Since managing a cluster consumes substantial energy, the cluster leader needs to be re-elected from time to time for load balancing. In hostile environments, it is critical to ensure the security of such leader election. This paper proposes an efficient, resilient, and fully distributed leader election protocol for sensor networks. It only uses efficient symmetric key operations and guarantees that (i) benign cluster members will elect the same leader as long as they are wellconnected, and (ii) attackers cannot impact the leader election process to increase or decrease the chance of a benign member being elected as a cluster leader. In addition, the proposed method can quickly recover from message loss or malicious attacks. The evaluation results also demonstrate the efficiency and effectiveness of this approach. I. INTRODUCTIONIn a densely-deployed sensor network, sensors are usually organized into clusters for efficiency and scalability purposes. Every cluster contains a number of sensors that are physically close to each other and is managed by a cluster leader to facilitate the network operation such as in-network aggregation and routing. For instance, in data aggregation, a cluster leader can serve as the aggregator that collects the sensing results from other cluster members, computes the aggregation result, and reports the result to the base station.Since managing a cluster consumes substantial energy, the cluster leaders need to be re-elected from time to time for load balancing. The leader election problem is to ensure that each cluster will have a suitable sensor node be selected as the cluster leader whenever needed. A common election metric to determine the new leader of a given cluster is the remaining energy on sensor nodes. For the sake of presentation, we call the values (on sensor nodes) used for election purpose as the election values. In the above case, the election values will be the remaining energy reported by sensor nodes.A number of secure clustering protocols have been proposed recently for organizing clusters in sensor networks [1], [2]. These protocols significantly improve the security of clustering in the presence of malicious attacks. However, they only focus on the formation of clusters and do not consider the security of cluster leader election. Without proper protection, an attacker can easily subvert the intended purpose of clustering. For example, if the protocol selects the node with the most remaining energy as the new cluster leader, an attacker can continuously hijack the leadership by always claiming more remaining energy than anyone else in the cluster.

On the data gathering capacity and latency in wireless sensor networks

“…More specifically, it is assumed that spatial TDMA is used at the MAC layer (which requires time synchronization), a weak form of locationawareness (nodes must know the cell to which they belong), and that leader nodes are selected in each cell. Indeed, time synchronization (see, e.g., [21]) , location-awareness (see, e.g., [14]), and leader election mechanisms (see, e.g., [3]) are quite typical building blocks of a wireless sensor network used, e.g., for environmental monitoring, so we believe that the proposed data gathering scheme can potentially be implemented in several wireless sensor network application scenarios.…”

“…Notice that although Zheng and Barton's results are based on less demanding assumptions than [3], they still build upon exploitation of advanced PHY layer techniques (cooperative communication), that might be difficult to use in a wireless sensor network.…”

“…To ensure collision-free delivery of data to the sink and optimal latency (in case minimal cell connectivity is satisfied), data is conveyed to the sink through a cell-based shortest path tree, called the data gathering tree, where the transmission time (slot) for each data source is appropriately computed 3 . An example of data gathering tree is reported in Figure 4: for each source node v, a minimal length cell path to sink node 3 A similar construction and transmission time computation has been used in the broadcast scheme for arbitrary number of sources presented in [16].…”

“…An example of data gathering tree is reported in Figure 4: for each source node v, a minimal length cell path to sink node 3 A similar construction and transmission time computation has been used in the broadcast scheme for arbitrary number of sources presented in [16]. s, denoted P(v, s), is randomly selected as data forwarding path 4 .…”

Securing Cluster Head Elections in Wireless Sensor Networks

“…The scheme's ability highly relies on their ID assignment scheme which tightly couples a node's ID, its commitments, and its polynomial shares [3]. In this scheme, members which do not distribute a participation message for CH election or explicitly distribute a non-participation message are removed from the CH candidates.…”