MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106684 A) filed by Soumya George; Dayananda Sagar Academy Of Technology And Management; Akshobhya K S; Nithin T A; Mansoor S; and Kushal Abhishek on September 04, 2026, for Ai Based Upi Fraud Detection And Behavioural Risk Scoring Using Ml.
Inventors include Soumya George; Dayananda Sagar Academy Of Technology And Management; Akshobhya K S; Nithin T A; Mansoor S; and Kushal Abhishek.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: Digital payment adoption has increased the volume and speed of UPI transactions, while fraudulent schemes such as fake login attempts, identity misuse, unauthorized transactions and other suspicious activities have also increased. The proposed AI-based UPI fraud detection system combines transaction analysis, behavioural assessment, device and network information, machine learning classification, anomaly detection and dynamic risk scoring for real-time fraud identification. UPI transaction data, user profile information, device/network data and location information are collected and processed through validation, cleaning, normalization, deduplication and missing-value handling. Behavioural features such as transaction frequency, spending behaviour, transaction patterns, location changes and device usage are compared with historical activity to identify deviations. Logistic Regression, Random Forest and XGBoost are used as classification models, while Isolation Forest and Autoencoders support anomaly detection. A risk-scoring engine assigns a fraud probability/risk level and an alert and decision engine can support actions such as transaction blocking, step-up authentication or manual review. The architecture also provides storage, monitoring, dashboards and feedback-based retraining to improve adaptability and reduce false positives.
Disclaimer: Curated by HT Syndication.
