No abstract

Cognitive radio is the next-generation wireless communication network that improves the efficiency of the radio spectrum through exploitation of underutilized licensed spectrum (or white spaces). This paper applies a reinforcement learning-based Trust and Reputation Management (TRM) scheme to cluster-based routing and shows network performance enhancement, including throughput and rewards. Generally speaking, clustering forms logical groups of nodes throughout the entire network, and routing establishes routes on the underlying clustered network which is distributed in nature. Each cluster is comprised of a clusterhead (or the leader of the cluster) and member nodes. TRM is applied as a security measure to allow each node to determine credibility of its neighbouring nodes. Reinforcement Learning (RL) is applied to keep track of the credibility level of a node, and it provides reward based on a node's behaviour; subsequently the reward is applied to select clusterheads. The selection of trusted nodes as clusterheads has been a problem that is of great significance due to the important role played by clusterheads as the local point of process for various applications such as channel sensing and routing. Our simulation results show that the RL-based TRM approach applied to clusterhead selection helps to reduce the effects of attacks from malicious nodes, and this has been shown to increase average throughput and reward rate, as well as to reduce changes of clusterhead in a cluster.

No abstract