Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{199052,
        author = {Nivashini s and Prof Dharmaraj TB and Manimegala s and Kirubadharshini R and Navaneethan C},
        title = {UPI Fraud Detection System Using Machine Learning Model},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {14167-14174},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199052},
        abstract = {Digital payment platforms, especially the Unified Payments Interface (UPI), have quickly changed how we handle money, making transfers instant and easy. But this widespread use also led to a surge in fraud, including phishing, social engineering, and unauthorised transactions. Spotting these fraudulent activities in real-time is a significant challenge for financial institutions and payment providers. This project presents a web-based fraud detection system, “UPI Fraud Detection using Machine Learning”, designed to flag suspicious transactions using data analysis and machine learning.
Built with the Python Flask framework, the system's backend manages user interactions, model predictions, and transaction analysis. It uses an XGBoost machine learning model, trained to classify transactions as legitimate or fraudulent by analysing specific behavioural and transaction features. This trained model is saved as a serialised Joblib file and loaded into the Flask app at runtime. The application collects five features from the user through a guided web interface: transaction amount compared to sender history, PIN entry speed, session duration, handle-to-description consistency, and app switching frequency. These features capture user behaviour and transaction details, helping distinguish normal activity from potential fraud.
The system's interface guides users through several pages, collecting each feature value and storing it with session management. After gathering all inputs, the application processes the data and sends it to the trained model for a prediction. The model then calculates the probability of fraud using its predict_proba() method, classifying the transaction as FRAUD or LEGITIMATE depending on that probability. The final result page shows the fraud probability, the prediction (fraud or legitimate), and the feature values that led to that decision. This makes the analysis clear.
Combining machine learning with a web-based interface, this system offers an effective way to detect fraudulent UPI transactions. It helps quickly identify suspicious patterns, helping financial platforms prevent fraud-related losses. This system demonstrates the power of merging machine learning models with web technologies to build effective, practical fraud detection solutions for digital payment environments.},
        keywords = {},
        month = {April},
        }