Enhancing Security and Privacy in the Internet of Things

Abstract

The Internet of Things (IoT) is revolutionizing the way we interact with the world around us, connecting everyday objects to the Internet for collecting and exchanging data. It o↵ers immense potential for enhancing e ciency, convenience, and technological innovation. However, the rapid expansion of IoT devices also introduces significant security and privacy challenges. These challenges arise because of the diverse and ubiquitous nature of IoT devices with limited processing capabilities and open networks, making them vulnerable to cyberattacks, data breaches, and unauthorized access. This thesis is based on novel security and privacy solutions for the unique constraints and requirements of the IoT ecosystem. First, we explore the application of blockchain technology as a foundational measure to secure and decentralize IoT systems. Recognizing the scalability challenges inherent in existing blockchain solutions when faced with the voluminous data generated by IoT devices, we propose an architectural framework that employs sidechains and o✏ine data storage. Extending this work further, we propose a hierarchical blockchain-based framework specifically designed for the healthcare sector. This framework addresses the issue of securely storing and sharing sensitive medical data without compromising its integrity and privacy. It optimizes data storage solutions and employs fog nodes for computational services to ensure e cient data management. Next, we explore the privacy aspect of IoT by leveraging Federated Learning (FL) to protect user data. We present an FL-based framework that enables collaborative training of a deep neural network on decentralized data, e↵ectively mitigating privacy risks associated with data sharing. We further expand our discussion to decentralize the FL framework by eliminating the need for a central server for data aggregation. Lastly, we develop a sustainable and secure framework for IoT-based applications by integrating the characteristics of both blockchain and FL. This mechanism enables the dynamic optimization of blockchain performance parameters by training a decentralized FL model, thereby preserving the distributed nature of blockchain while enhancing its e ciency. Through theoretical analysis and practical implementation, we validate the feasibility and performance of proposed frameworks for di↵erent applications of IoT.