Blockchain-Integrated Machine Learning for Financial Data Provenance Verification: Ensuring Transaction Authenticity and Traceability Across Multi-Institutional Financial Systems.

Abstract

Modern financial networks span multiple institutions, which creates a tough double problem. They have to spot fraudulent transactions early while also making sure the financial records themselves stay authentic, traceable, and locked down against sneaky changes over time. Normal fraud detection systems that rely purely on machine learning are great at flagging strange patterns, but they cannot actually prove that the transaction data has not been messed with after the fact. This study builds and tests a system that connects a machine learning model with a blockchain layer to fix that gap. By combining real-time fraud spotting with strict history tracking, the setup covers both bases. The framework uses a lightweight, permissioned blockchain simulation to cryptographically track where transactions come from and where they go. On top of that, it runs several supervised learning models, specifically Logistic Regression, Decision Tree, Random Forest, XGBoost, LightGBM, and CatBoost, trained on the standard IEEE-CIS Fraud Detection dataset. To see how it handles real pressure, the system was tested inside a simulated multi-bank network where realistic data tampering attacks were thrown at it. The performance was analyzed by putting three setups against each other: a machine learning system on its own, a blockchain system on its own, and the combined blockchain-machine learning setup. The tests show that the blockchain layer successfully flags unauthorized data changes that a normal machine learning model completely misses. When it comes to sorting the fraudulent activities from the legitimate ones, the gradient boosting models perform the best. Ultimately, the combined system offers much tighter security because it pairs predictive analytics with cryptographic proof of where the data has been. The results show that blockchain and machine learning fix two different sides of the same problem. Blockchain ensures the transaction history is permanent and easy to follow, while machine learning catches the shady behavior itself. This work offers a practical blueprint for building smart financial networks that keep data honest and transparent across different institutions.