Yonsei University

Ju Sik Yun

Jong-Moon Chung

Yun Yeong Goh

An apparatus for managing delegation consensus of a blockchain network, obtains state information including trust of each delegated node in the current epoch of the blockchain network, determines an action including a delegation rate by performing a neural network operation on the state information, creates, according to the determined action, a local trust opinion (hereinafter, LTO) matrix representing a mutual subjective evaluation of each delegated node, and selects, based on the trust calculated using the LTO matrix, K delegated nodes according to the determined delegation rate among N nodes.

Apparatus and method for managing trust-based delegation consensus of blockchain network using deep reinforcement learning

A shard distribution apparatus and method are disclosed. The shard distribution apparatus can include a trust calculation unit configured to obtain a trust value for each of a multiple number of nodes in a blockchain network by receiving a consensus result from each of the multiple nodes, where the consensus result may be obtained by aggregating the result of verifying the validity of a block by each of the multiple nodes and the results of verifying the validity of the block received from other nodes; and a shard distribution unit configured to distribute the multiple nodes to a multiple number of shards by calculating a shard trust value, which is represented by a sum of trust values of nodes distributed to each of the multiple shards, and distributing the multiple nodes such that deviations of the calculated shard trust values are the smallest.

Trust-based shard distribution apparatus and method for fault tolerant blockchain networks

An apparatus for managing sharded blockchain network obtains state information including an inter-node data transmission rate and computing resources of each node in the current epoch of a shard-based blockchain network in which N nodes are distributed to at least one shard, a consensus process record of each node in each shard and a ratio of malicious nodes, estimates at least one action each including block size, block interval and number of shards corresponding to the state information by using an artificial neural network in which a pattern estimation method has been previously learned, selects one action among the estimated at least one action and applies the one action to the shard-based blockchain network.

Growing community of inventors

Seoul, South Korea

Ju Sik Yun