ELM-based data distribution model in ElasticChain

Blockchain technology is becoming familiar to the public, along with the widespread use of cryptocurrency. The blockchain protocol requires that full nodes need to save the complete blockchain data, which limits the joining of resource-constrained nodes. A small number of full nodes will reduce the decentralize and security of system. Elasticchain was proposed in 2018 to solve this problem by saving fragments of the entire blockchain in reliable nodes. However, Elasticchain does not give an effective method to evaluate the reliability of nodes. If the fragmented data is stored in unreliable nodes, such as malicious tampering, are often not online or the latency is too high, the security of blockchain system will be seriously impacted. Therefore, in this paper, we propose an ELM-based method to comprehensively evaluate node reliability, and the blockchain system distributes the fragmented data to reliable nodes for storage. In the new method, ELM is used as a classifier to select reliable nodes because the ELM has a higher performance of training and classification compared to other machine models. Moreover, in ELM classifier five novel evaluation features are considered: the security, the trustworthiness, the activeness, the stability and the communication costs. Finally, the experimental results on synthetic data demonstrate the accuracy and efficiency of the optimized data distribution model.