A Blockchain-Integrated Incremental Learning Framework for Adaptive Fraud Detection in Financial Transactions

Abstract

The rapid growth of online financial transactions has significantly increased the risk of fraud, creating critical challenges for modern financial systems. With the expansion of Internet of Things (IoT) environments and the continuous rise in transaction data volume, real-time fraud detection becomes difficult due to large-scale data streams, class imbalance, and evolving fraud patterns. Traditional centralized machine learning (ML) approaches often suffer from limited scalability, delayed model updates, and reduced adaptability to dynamic data, leading to performance degradation over time. To address these limitations, this work proposes a blockchain-integrated incremental ML framework for fraud detection. The system evaluates multiple ML algorithms, including Passive Aggressive Classifier (PAC), Stochastic Gradient Descent (SGD) Classifier, Perceptron, Naïve Bayes (NB), and Light Gradient Boosting Machine (LGBM), to compare performance. Among them, the incrementally trained SGD classifier is selected as the final model due to its efficiency, scalability, and strong capability in handling streaming data. To mitigate class imbalance and improve detection accuracy, Synthetic Minority Oversampling Technique (SMOTE) is employed. Furthermore, blockchain technology is integrated to securely store model parameters, ensuring data integrity, transparency, and decentralized control. The system is implemented using the Flask web framework, enabling real-time fraud prediction through an interactive interface. The proposed approach improves detection accuracy, reduces training time via incremental learning, and enhances security with tamper-proof storage, making it a scalable and reliable solution for financial and IoT-based fraud detection systems.