Combınatorıal Solutıons For Consensus Algorıthms And Blockchaın Shardıng

Abstract

The scalability problem in blockchain technology seems to be the essential issue to be solved. It is known that the choice of a compromised algorithm is critical for the practical solution of this important problem. Usually, Byzantine Fault Tolerance (BFT) methods based on the public blockchain networks have been most widely applied to solve scalability. In this thesis, we formulate two possible cases to scale the blockchain. Instead of the frequently used proof-of-work or stake methods to form the consensus committee, allowing BFT-based methods, we propose a new model. This new model calculates the reputation value for the nodes that want to join the leader (trust) committee using particle swarm optimization (PSO). It is a computational method for optimizing a problem by improving a candidate solution against a specified quality metric. It solves the problem by populating the search space, so-called particles and moving these particles around according to a simple mathematical formula over the particle's position and velocity. To discard the misbehaving nodes from the trust leader committee, new nodes with high reputation values are selected. Since this study focuses on creating the consensus committee, a simulation test the proposed model more effectively. The results show that the proposed model successfully selects the nodes with high confidence to the consensus committee instead of the malicious nodes. To select an updated trustworthy committee and then allow all network users to join at any time to protect the blockchain network's security is in general insufficient. However, suspicious nodes must be avoided at all costs. We utilize a straightforward strategy inspired by bio-dynamic systems to deflect the trust committee's focus from the assaulting nodes. Removing poorly-tailored nodes increases the selection of honest nodes or participants. We propose an unsupervised machine learning to solve the current challenge by applying a Grey Wolf Optimization (GWO) technique. In addition, blockchain studies have recently been splitting the blockchain to address the scalability problem focused on sharding. Sharding is a helpful technique for exploring fundamental computational challenges in blockchain technology, such as consensus, Byzantine fault tolerance, and self-stabilization. The sharding method creates a small, segmented blockchain network. Rather than creating a more extensive network, networks with fewer nodes are established. Additionally, successful sharding can be applied to various areas, resulting in significantly speedier processes. Our solution will give a safe and dependable use of blockchain components by analyzing the system and fitting the shard size using the Topological Data Analysis (TDA) with the help of an unsupervised machine learning technique. In order to achieve our goal, the Linear Programming Problem (LPP) is constructed and solved using the Dual-Simplex approach to determine the best shard size. Additionally, we segmented the blockchain network using our system. The test results show that reputation values boosted the parties' reliability. Then the likelihood of any piece collapsing and harming the entire blockchain decreases.