Content area
Blockchain technology has significantly evolved, notably through the adoption of sharding techniques to enhance scalability and manage growing transaction demands efficiently. Sharding divides the blockchain network into smaller, more manageable segments, or shards, which operate parallel transactions, thereby boosting throughput. However, conventional sharding solutions often encounter challenges related to node distribution and transaction efficiency, particularly when managing cross-shard transactions which can significantly degrade network performance and increase transaction costs.
In this thesis, Blockchain Sharding bottlenecks are addressed by introducing novel methods to optimize sharding schemes for enhanced efficiency and security. Firstly, a community detectionbased sharding scheme are validated using over one million public Ethereum transactions. This scheme dramatically decreases the ratio of cross-shard transactions from 80% to 20%, significantly lower than that of baseline random sharding methods. By clustering transactional communities, community detection-based sharding approach promotes intra-shard transactions, which substantially reduces transaction fees by 50%, thereby enhancing blockchain scalability and making the ecosystem more budget friendly. Furthermore, a novel Trust-based and Deep Reinforcement Learning-driven (TbDd) framework are introduced to mitigate collusion risks and dynamically adjust node allocation, thus enhancing throughput while maintaining network security. The TbDd framework features a comprehensive trust evaluation system that identifies node behaviors and performs targeted resharding to address potential threats. Designed to increase tolerance for dishonest nodes, optimize node movement frequency, and ensure equitable distribution across shards, this framework effectively balances sharding risks. Extensive testing shows that TbDd surpasses traditional random, community, and trust-based methods in maintaining shard risk equilibrium and reducing cross-shard transactions. Lastly, an advanced Overlapping Sharding with xPBFT Consensus Mechanism are proposed, which simplifies cross shard transactions by treating them as intra-shard processes. This approach reduces latency by up to 40% and strengthens security, offering a scalable solution for decentralized applications. These contributions advance the efficiency, security, and scalability of sharded blockchain systems, providing a robust foundation for future developments in blockchain technology.