Reinforcement Learning-Based Trust and Reputation Model for Spectrum Leasing in Cognitive Radio Networks

Cognitive Radio (CR), which is the next generation wireless communication system, enables unlicensed users or Secondary Users (SUs) to exploit underutilized spectrum (called white spaces) owned by the licensed users or Primary Users (PUs) so that bandwidth availability improves at the SUs, which helps to improve the overall spectrum utilization. Collaboration, which has been adopted in various schemes such distributed channel sensing and channel access, is an intrinsic characteristic of CR to improve network performance. However, the requirement to collaborate has inevitably open doors to various forms of attacks by malicious SUs, and this can be addressed using Trust and Reputation Management (TRM). Generally speaking, TRM detects malicious SUs including honest SUs that turn malicious. To achieve a more efficient detection, we advocate the use of Reinforcement Learning (RL), which is known to be flexible and adaptable to the changes in operating environment in order to achieve optimal network performance. Its ability to learn and re-learn throughout the duration of its existence provides intelligence to the proposed TRM model, and so the focus on RL-based TRM model in this paper. Our preliminary results show that the detection performance of RL-based TRM model has an improvement of 15% over the traditional TRM in a centralized cognitive radio network. The investigation in the paper serves as an important foundation for future work in this research field.