Predicting Node Colocality for an Auto-Scaling Multi-Tiered Node-Organizing Blockchain Architecture for IoT

Current node-organizing blockchain models for IoT lack the required scalability to handle more IoT devices, mechanisms to balance decentralization versus node overhead, as well as node proximity considerations to reduce consensus time. In previous studies, we proposed a multi-tiered node-organizing model, based on a forest to improve blockchain scalability and efficiency by clustering all IoT nodes throughout the tiers, limiting those involved in consensus. At the heart of the model lies a protocol to accelerate consensus, a mechanism to efficiently perform node-lookup and message-routing, a dynamic algorithm to balance overhead and decentralization, as well as a node colocality concept that proved to further reduce consensus time by about 17% in about 90% of the runs, based on our previous performance analysis. In this study, we propose a model to predict future node interactions using LSTM and Diffusion Models with the aim to accordingly optimize node organization, increasing the percentage of colocality, hence further reducing consensus time.