Quantique Hackathon | QuantumShield – UPI Fraud Defence | UC-021 | Team KMPD

This video is our Round 1 presentation for the **Quantique Hackathon 2026**. Our project, **QuantumShield**, focuses on detecting emerging UPI fraud when there is very little labelled data available. We specifically looked at a rapid device-hopping pattern, where an account starts using multiple devices within a short period of time. In this project, we compare a *quantum-kernel approach using Qiskit* with a classical **XGBoost model**. We tested both approaches with different amounts of labelled data, starting from 0 labels and gradually increasing up to 2,000 labels. The main idea of our work is to see whether a quantum-kernel model can provide a useful signal during the early stage, when there is not enough data to train a conventional model effectively. Our benchmark showed an interesting trade-off between the two approaches. The quantum model showed a stronger signal at the cold-start point, while XGBoost performed better as more labelled data became available. We have also considered the false-positive rate and latency in our comparison. This work is a controlled benchmark and does not claim general quantum advantage or production readiness. *Team KMPD* B. Lakshmi Surya Teja Sri Mahalakshmi Bhavaganya Sri Raajitha *Use Case:* UC-021 – Real-Time UPI & Digital-Payments Fraud Defence *Event:* Quantique Hackathon 2026 #QuantumShield #Qiskit #QuantumComputing #UPI #FraudDetection #TeamKMPD