Scholarlink

<p>Financial fraud, particularly cheque fraud, presents a major challenge to banking security, necessitating advanced detection and prevention mechanisms. Traditional systems often struggle to efficiently manage and authenticate large volumes of transactional data, leading to delays and inaccuracies. To address these limitations, this study proposes a novel framework, AI-MCAGCN-B-DIFCS<b>,</b> which integrates a Multi-Component Attention Graph Convolutional Neural Network (MCAGCN) with blockchain technology for secure and precise fraud detection. Transactional data are sourced from the World Bank Financial Sector Data for trend analysis and the IEEE-CIS Fraud Detection Datasetfor time-stamped, real-time cheque simulations. The structure it makes use of a Fair Proof-of-Reputation (PoR) blockchain consensus mechanism to authenticate cheque transactions, ensuring tamper-proof, transparent, and reliable data recording across domestic and international financial institutions. Once validated, cheque data are analyzed using the MCAGCN model, which leverages graph-based structures and attention mechanisms to capture complex relationships, patterns, and anomalies within the transactions, enabling accurate differentiation between legitimate and fraudulent cheques. The Red-Billed Blue Magpie Optimizer (RBMO) is utilized to fine-tune model parameters, enhancing both predictive accuracy and computational efficiency. Implementation in Python demonstrates the framework’s effectiveness, achieving superior performance compared to conventional models such as ANN, RNN, and CNN. Specifically, the AI-MCAGCN-B-DIFCS model shows 29–30% higher accuracy and 17–36% reduced computation time, along with improvements in precision, recall, F1-score, RMSE, and specificity. These results highlight the combined advantages of AI-based graph learning and blockchain integration, providing a robust, transparent, and scalable solution for real-time cheque fraud detection and prevention in modern banking environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced AI methods for cheque book scam prevention using block chain and multi-component attention graph convolutional networks

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • M. Srinivasa Varma,
  • G. Pardhasaradhi Varma,
  • I. Hemalatha

摘要

Financial fraud, particularly cheque fraud, presents a major challenge to banking security, necessitating advanced detection and prevention mechanisms. Traditional systems often struggle to efficiently manage and authenticate large volumes of transactional data, leading to delays and inaccuracies. To address these limitations, this study proposes a novel framework, AI-MCAGCN-B-DIFCS, which integrates a Multi-Component Attention Graph Convolutional Neural Network (MCAGCN) with blockchain technology for secure and precise fraud detection. Transactional data are sourced from the World Bank Financial Sector Data for trend analysis and the IEEE-CIS Fraud Detection Datasetfor time-stamped, real-time cheque simulations. The structure it makes use of a Fair Proof-of-Reputation (PoR) blockchain consensus mechanism to authenticate cheque transactions, ensuring tamper-proof, transparent, and reliable data recording across domestic and international financial institutions. Once validated, cheque data are analyzed using the MCAGCN model, which leverages graph-based structures and attention mechanisms to capture complex relationships, patterns, and anomalies within the transactions, enabling accurate differentiation between legitimate and fraudulent cheques. The Red-Billed Blue Magpie Optimizer (RBMO) is utilized to fine-tune model parameters, enhancing both predictive accuracy and computational efficiency. Implementation in Python demonstrates the framework’s effectiveness, achieving superior performance compared to conventional models such as ANN, RNN, and CNN. Specifically, the AI-MCAGCN-B-DIFCS model shows 29–30% higher accuracy and 17–36% reduced computation time, along with improvements in precision, recall, F1-score, RMSE, and specificity. These results highlight the combined advantages of AI-based graph learning and blockchain integration, providing a robust, transparent, and scalable solution for real-time cheque fraud detection and prevention in modern banking environments.