Blockchain-Enabled Federated Reinforcement Learning (B-FRL) Model for Privacy Preservation Service in IoT Systems

Tanweer Alam, Ruchi Gupta, Arif Ullah, Shamimul Qamar

Abstract: The rapid explosion of Internet of Things (IoT) devices has resulted in extraordinary data generation, necessitating advanced methods for data processing and privacy preservation. The IoT seeks to enable seamless communication between any object or device and any service provider without any time constraints. On the other hand, IoT devices might expose users to numerous privacy and security risks. The primary goal of privacy-preserving methods is to ensure the confidentiality and integrity of all data transfers. Blockchain technology ensures the authentication and prevention of any duplication of sensor data, hence removing any inaccurate information. Sensors can employ blockchain technology to communicate data autonomously, eliminating reliance on a trusted intermediary. The IoT devices collect and use personal data, giving rise to privacy concerns. Users may need to comprehend the full scope of data collected and its intended utilisation. The user may also be unaware that their information could be revealed to third parties. The Federated reinforcement learning (FRL) framework provides effective methods for guaranteeing the security of data and models in IoT systems. Blockchain ensures data security during data transmission among IoT devices. FRL integrates federated learning (FL) and reinforcement learning (RL) that guarantees the privacy of both the models and the data. FRL addresses issues with agents, safeguards privacy, ensures the confidentiality of agents, and produces global models. Local models are trained on separate devices to fulfil data privacy goals. In this paper, the authors have explored the B-FRL model in IoT systems to provide privacy protection services.