Threat Guard AI: A Multi-Domain System for Intelligent Fraud Detection Using Machine Learning

Abstract

In the era of rapid digital transformation, cyber threats such as phishing, QR code manipulation, UPI fraud, and PDF malware have become increasingly sophisticated. Threat Guard AI is a comprehensive, multi-domain fraud detection system designed to counter these threats using artificial intelligence and machine learning. The platform integrates specialized modules to analyze UPI transactions, QR codes, PDF documents, and phishing links. Through a user-centric Streamlit interface and a robust Flask backend, it provides real-time analysis, prediction, and alerts. Each module is powered by trained models using Scikit-learn and OpenCV, with MongoDB for scalable data storage. Threat Guard AI not only enhances the precision of fraud detection but also ensures actionable insights, unified visualization, and extensibility to new threat domains. This paper presents the architecture, implementation methodology, and system performance analysis of Threat Guard AI, demonstrating its capability to provide intelligent, real-time digital threat mitigation.

Introduction

Context:
With the rise of digital transactions and communication, cyber threats and financial fraud have become more sophisticated and harder to detect. New payment methods like APIs, digital wallets, and UPI have introduced vulnerabilities exploited by fake QR codes, malware, and phishing attacks.

Problem:
Traditional security systems are reactive and isolated, failing to address the increasing diversity and complexity of fraud across platforms.

Solution – Threat Guard AI:
An integrated, AI-powered fraud detection platform that monitors four major threat types:

• UPI fraud (unauthorized digital payments)

• QR code fraud (malicious QR codes)

• PDF malware (infected document files)

• Phishing attacks (fraudulent attempts to steal credentials)

System Architecture:

• Frontend: Built with Streamlit for an interactive dashboard, offering modules dedicated to each fraud type and real-time alerts.

• Backend: Flask API server managing ML model execution, data processing, and communication with MongoDB and Firebase databases.

• Database: MongoDB stores structured and unstructured data related to transactions, QR codes, PDFs, and phishing URLs.

Methodology:

• Multi-layered detection combining rule-based logic and machine learning (Random Forest classifiers, NLP for phishing URLs, image processing for QR codes and PDFs).

• Real-time monitoring with low latency and scalable, secure architecture.

Implementation Details:

• Modular design with separate detectors for each fraud type.

• Models trained on historical fraud data, evaluated for accuracy and recall.

• User-friendly interface for both end-users and security professionals.

Testing:
Validated through multiple test cases (user registration, malicious PDF detection, QR code scanning, UPI fraud alerts, phishing URL identification), all successfully passing.

Results:
The platform offers a visually engaging dashboard displaying system status and interactive modules for file threat analysis, QR scanning, fraud detection, and phishing protection. The UPI module analyzes transaction parameters to flag suspicious activities with high confidence.

Conclusion

The above-discussed multi-domain fraud detection system is a huge leap toward the modernization of fraud prevention. By leveraging the power of AI and a unified detection platform, it fills critical gaps in traditional systems. The design ensures real-time threat analysis, modular scalability, and data-driven predictions. Continuous testing, improvement, and extension into cloud infrastructure will further amplify its impact on the digital security landscape.

References

[1] Ume Zara et al., “Phishing Website Detection Using Deep Learning Models,” IEEE Access, 2024. [2] S. Jagadeesan et al., “UPI Fraud Detection Using Machine Learning,” International Journal of Security Studies, 2025. [3] J. Seetha et al., “QR Code Recognition Based on Image Processing,” Springer LNCS, 2024. [4] I. Ahmed et al., “DeepLedger: A Blockchain-Integrated Deep Learning Model for Malware,” ACM CCS, 2023. [5] R. Kumar et al., “A Survey on Fraud Detection Mechanisms in Mobile Payment Systems,” IEEE Transactions on Consumer Electronics, vol. 68, no. 2, pp. 234-245, Apr. 2022. [6] S. Gupta and N. Singh, “Phishing Detection Using Hybrid Deep Learning Approach,” Journal of Cybersecurity and Privacy, vol. 6, no. 3, pp. 98-112, Mar. 2023. [7] M. Patel and T. Singh, “QR Code-Based Authentication: Security Risks and Mitigation,” International Journal of Computer Science and Security, vol. 17, no. 4, pp. 256-268, Dec. 2023. [8] P. V. Narayanan et al., “AI-Driven UPI Transaction Fraud Detection in Real-Time,” International Journal of Information Security, vol. 19, no. 1, pp. 45-57, Jan. 2024. [9] R. Shah and A. Agarwal, “Exploring Malware Detection in PDFs Using AI Models,” Proceedings of the International Conference on Digital Forensics and Cybersecurity, pp. 180-190, 2024. [10] H. Joshi et al., “Application of Blockchain for Secure Payment Transactions,” IEEE Blockchain Conference, pp. 25-30, 2023.

Copyright

Copyright © 2025 Pranjal Bhinge, Ranjeeta Chikkaparappa, Shruti Basarikatti, Rohan Pawar. This 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.