Adaptif Öğrenme Tabanlı Güven Değeri Kullanılarak Blokzincirin Ölçeklenmesi

Abstract

Scalability has become a challenging problem for blockchain technology. Consensus algorithm selection is critical for the practical solution of the scalability problem. Byzantine Fault Tolerance (BFT) based methods have been applied most commonly to increase scalability. We propose a new model for creating consensus committee which is not using Proof of Work (PoW) so that BFT-based methods could be used in public blockchain networks. In our model, we use an online, decision-theoretic, unsupervised learning algorithm which is called the adaptive hedge method [1]. For nodes wishing to join the consensus committee, the reputation value is calculated and nodes with a high reputation value are selected to the consensus committee to reduce the probability that the nodes in the consensus committee are harmful. Since this study focused on establishing a consensus committee, simulation of the blockchain network was used to test the proposed model more effectively. The test results show that the proposed model (a new approach that uses machine learning in the creation of a consensus committee) has successfully selected nodes with high reputation in the consensus committee. In addition, blockchain studies have recently focused on sharding the blockchain for solving the scalability problem. Sharding method divides the blockchain network into small pieces. Networks with fewer nodes are created instead of a more extensive network. Therefore, it becomes more important for every node in the network to be reliable. Using adaptive learning-based methods for this process will contribute to the safe and reliable use of blockchain pieces. The probability of each piece breaking and affecting the entire blockchain will be reduced. We used our model to shard the blockchain network and we see that using reputation value increases shard's reliability in our test results.