A novel approach to enhance fault tolerance in vote-based blockchain consensus algorithm using machine learning mechanism

Abstract The blockchain technology has shown promise of transforming a variety of industries, it has captured significant interest for its ability to revolutionize multiple sectors through the facilitation of decentralized and secure transactions. Peer-to-peer architecture of blockchain provides strong security and trust-oriented guarantees, such as immutability, verifiability and decentralization. Nonetheless, it remains a considerable challenge to ensure the safety and efficiency of blockchain networks. Through consensus algorithms, existing blockchain can still maintain its integrity and safety; however, existing consensus tools face hurdles in addressing security threats and attacks. We need an advance method to improve security without slowing down the blockchain network. This paper proposes an advanced technique used to identify anomalous nodes in a blockchain network through the use of a machine learning technique that enhances the security and reliability of new node addition or a new transaction. A novel model for a blockchain network based on a vote-based consensus algorithm is presented, which enhances its fault-tolerant. The proposed method will monitor the network pattern by implementing machine learning algorithms to detect anomalies and prevent fraud. We will be making use of the vote-based approach, which will be performed by the nodes or the validators to check whether to approve or validate the transaction. Through simulations and experiments, the effectiveness of the approach can be evaluated. The proposed model achieved an impressive F1 score of 0.68, precision of 0.70, and recall of 0.66 in the case of the Elliptic Bitcoin Transaction dataset in the context of unsupervised learning outperforming the work presented in the literatures. Further, the inclusion of the machine learning-based node filtering approach to the consensus process resulted in the reduction of the consensus latency of PBFT/DBFT by approximately 15%. The result shows that adding the functionality would enhance blockchain security, allowing for additional advancements.