Abstract
Blockchain is a technology for recording information across distributed networks of computers, widely used in fields as wide-ranging as cryptocurrency, smart contracts, logistics and supply chain management, healthcare, insurance, copyright protection, and more. Its core operation is decentralised consensus that enables data and transaction validation to provide security, reliability and accuracy without a central authority. However, existing distributed consensus protocols suffer from several limitations, including node failures, excessive power consumption, slow confirmation time, poor scalability, and susceptibility to collusion. In this thesis, we present the Blockchain Epidemic Consensus Protocol (BECP), a fully decentralised consensus protocol designed for extreme-scale blockchain systems. BECP leverages epidemic protocols and confers advantages such as obviating the need for fixed validators, logarithmic diffusion, probabilistic convergence guarantees, and efficient network resource usage. It is tolerant to node failures, message delays and network dynamics such as nodes joining and leaving. Experimental comparisons with the classical protocols such as PAXOS, RAFT, and Practical Byzantine Fault Tolerance (PBFT), as well as epidemic-based protocols such as Avalanche and Snowman, demonstrate the improvements that BECP achieves concerning throughput, network overhead, and scalability. Specifically, BECP outperforms the epidemic-based decentralised protocols Avalanche and Snowman in terms of message overheads and average consensus time. Additionally, BECP results in a large reduction of the message overheads between nodes and maintains performance even in the presence of node failure and late message delivery. These results confirm the robustness and efficiency of BECP, establishing it as a milestone in decentralised consensus models.