Authors: S. Kushala Shree, Tanishqa, Akanksha P. Bhat, M. R. Prathima, Suzzaine, Jashwanth M. K
Abstract: Unified Payments Interface (UPI) has made digital payments convenient by allowing users to transfer money through supported payment applications. QR codes simplify this process because merchant details can be encoded in a pattern that a phone camera can read. However, a QR code can also be replaced, covered, altered, or used to direct a user towards an unintended payment destination. A person may not be able to identify the destination by looking at the QR image alone. This creates a need for an additional warning layer that checks the decoded information before a user proceeds. This paper proposes an AI-assisted UPI QR-code tampering detection system. The proposed workflow captures or uploads a QR image, decodes its payload, checks payment-related fields, extracts suspicious indicators, and uses a machine-learning classifier together with clear security rules to produce a risk category. The output may be shown as “Lower Risk,” “Needs Verification,” or “High Risk,” with a short explanation such as a mismatch between the expected merchant and decoded payee. The system is intended as a decision-support tool rather than a replacement for a bank or UPI application’s security controls. The study discusses the problem, related work, system architecture, data requirements, feature extraction, model selection, evaluation measures, limitations, and future scope. Since no validated dataset or live payment integration is supplied for this paper, the proposed system is not presented as a tested product. A prototype should be evaluated on representative, legally obtained data before any real-world deployment.
DOI: https://doi.org/10.5281/zenodo.23973008
