Hyperledger Fabric Graph Isomorphism Network for Conflict Transactions Detection in Multi-Version Concurrency Control

• Published in: IEEE access Vol. 13; pp. 131314 - 131333
• Main Authors: Alzahrani, Fawaz, Yazid Idris, Mohd, Fo'Ad Rohani, Mohd, Budiarto, Rahmat
• Format: Journal Article
• Language: English
• Published: IEEE 2025
• Subjects: Blockchains; Complexity theory; Concurrency control; convolutional neural networks; Data integrity; deep learning; Distributed ledger; Fabrics; graph neural networks; graph theory; hyperledger; machine learning; Proposals; Reliability; Scalability; System performance
• DOI: 10.1109/ACCESS.2025.3589165
• ISSN: 2169-3536

Multi-Version Concurrency Control (MVCC) is a critical mechanism in blockchain systems such as Hyperledger Fabric. MVCC is crucial for managing concurrent transactions ensuring integrity through versioning and timestamp techniques. Enhancing MVCC is essential due to the significant overhead involved in maintaining multiple versions of data and resolving conflict transactions, particularly in enterprise blockchain applications where data integrity and system reliability are of utmost importance. Conflict transactions are the predominant cause of transaction failures in Hyperledger Fabric where multiple transactions attempt to update data items based on outdated data versions. This study introduces the Hyperledger Fabric-based Modified Graph Isomorphism Network (HFGIN). Employing advanced graph neural networks, HFGIN utilizes node and edge data representations within a graph-based framework, which significantly increases the efficiency of detecting MVCC conflicts. The proposed model aims to detect conflict transactions at an earlier stage in Hyperledger Fabric blockchain systems. Evaluation results indicate that HFGIN substantially outperforms baseline models such as the Graph Convolutional Network (GCN) and Graph Isomorphism Network (GIN). HFGIN achieves a test accuracy of 82.20%, demonstrating a 14.67% improvement over GCN and a 4.12% improvement over GIN. Moreover, HFGIN records a precision of 86.06%, marking an 8.97% enhancement over GCN and 9.46% over GIN. It also shows a recall of 79.56%, which is 17.03% higher than GCN and 1.39% greater than GIN, besides an F1 score improvement of 23.37% over GCN and 5.27% over GIN. HFGIN also demonstrates computational efficiency with low inference latency and minimal resource usage and maintains scalability when tested under larger transaction loads. The model achieves these improvements while maintaining computational efficiency and system scalability by applying inference selectively and enabling parallel execution of non-conflicting transactions. The enhancements contribute to detection rate increase of conflict transactions over GCN and GIN in an earlier stage, demonstrating the model's potential to transform MVCC conflict management in high-demand blockchain environments such as banking operations.