摘要：

随着物联网规模化应用的不断落地，海量设备协同工作，产生了大量高价值数据。这些数据若得不到有效的安全保障，就容易遭受数据滥用、隐私泄露以及数据篡改等威胁。因此，去中心化、不可篡改、安全的区块链分片网络逐步取代传统的集中式网络，被应用到该场景中。然而，区块链分片网络受限于复杂环境以及高比例跨片协同事务。针对上述问题，提出一种面向6G物联网设备协同的区块链动态分片优化方案。设计分片系统架构，建立整个系统的吞吐量模型、安全模型以及时延模型。在此基础上，提出两阶段分片优化策略。第一阶段采用信誉分级分片策略筛选节点，第二阶段采用基于深度强化学习算法的动态分片策略，降低跨片协同事务比例，决策分片数量。两阶段的设计目的是在保证安全的情况下最大程度地提升整个系统的吞吐量。实验结果表明，在面向物联网设备协同的区块链分片场景下，相较于传统的基于单一的信誉分级分片策略、均匀分片策略或者随机分片策略的方案，所提方案在保证安全性的情况下，平均每轮减少50%以上的跨片协同事务比例，有效地提升了系统的吞吐量。

关键词: 物联网, 区块链分片, 委托拜占庭容错, 信誉值, 跨片协同, 深度强化学习

Abstract:

With the massive deployment of Internet of Things (IoT) applications, numerous devices work together to generate massive, high-value data. If the security of these data is not guaranteed, they are vulnerable to threats such as data abuse, privacy leakage, and data tampering. Therefore, a decentralized, immutable, and secure blockchain sharding network is applied in this scenario, gradually replacing the traditional centralized network. However, the performance of blockchain sharding network is limited by the high proportion of cross-shared collaborative transactions or complex environments. To solve these problems, a dynamic blockchain sharding optimization scheme for 6G IoT device collaboration is proposed. First, the sharding architecture of the system is designed, and throughput, security, and delay models of the system are established. Subsequently, a two-stage sharding optimization strategy is proposed. In the first stage, the nodes are screened using a reputation-based sharing and grading strategy. In the second stage, a dynamic sharding strategy based on Deep Reinforcement Learning(DRL) is used to reduce the proportion of cross-shard collaborative transactions and determine the number of shards. By combining these two stages, the throughput of the entire system can be improved while ensuring safety. The experimental results reveal that in the blockchain sharding scenario with the collaboration of IoT devices, the abovementioned scheme, compared with traditional schemes based on reputation-based grading sharding, uniform sharded, or random sharding strategies, can reduce cross-shard collaborative transaction proportions by over 50%. In this case, the throughput of the system improved.

Key words: Internet of Things(IoT), blockchain sharding, Delegated Byzantine Fault Tolerance(DBFT), reputation value, cross-shard collaboration, Deep Reinforcement Learning(DRL)