Efficient Consensus Architectures for Blockchain-Based IoT Systems

The Internet of Things (IoT) enables the interconnection of resource-constrained devices, or "things," through the Internet, and it is rapidly becoming one of the most popular technologies. However, attackers could take advantage of the centralized architecture to create disruptions in the network, alter collected data, or even affect the reputations of trusted devices. Since the blockchain ledger is decentralized, verifiable, and secure, it has been used in a variety of IoT application scenarios. As the nodes in the network do not trust each other, they must use a consensus protocol to ensure the validity and availability of accepted data items. In this dissertation, I propose a set of consensus protocols that address the main issues that practical Byzantine fault tolerance (PBFT) and other Byzantine fault-tolerant (BFT) protocols may face. The proposed BFT consensus protocols eliminate the dependence on a single leader node. In addition, they improve the system's scalability and reduce the long latency in the communication between the nodes and the clients, who may be located geographically far from the single leader node in leader-based BFT protocols. The first BFT-based protocol proposed in this dissertation eliminates the single leader node problem and can run multiple consensus rounds concurrently without a performance penalty. This eliminates the long latency and limited bandwidth that originate from the individual processes in most BFT-based protocols. This dissertation also introduces a spot reservation mechanism that processes multiple consecutive requests without contention, which improves the protocol throughput and scalability. The second proposed protocol provide a solution for blockchain-based IoT applications that may require wide geographic coverage. Finally, in the last proposed BFT-based consensus protocol, a mechanism is proposed that incentivizes the nodes to act honestly. To analyze and evaluate the performance of the proposed protocols, we develop detailed analytical models based on discrete-time Markov chain (DTMC) and queuing theory. The models reveal improvements in propagation delays, throughput, and mean response time compared with the standard single-stream PBFT protocol.