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{193755,
        author = {Pushpendra Neniwal and Dr. Vikas Kapoor},
        title = {SCARA: An Intelligent Agent Framework for Real-Time Fraud Detection in Indian Public Sector Supply Chains Using Synthetic Transaction Simulation and Large Language Model-Powered Risk Scoring},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {10},
        pages = {1279-1292},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=193755},
        abstract = {Fraudulent procurement practices within public sector supply chains represent a deeply entrenched and economically damaging challenge in large emerging economies, where the volume of transactions is enormous and oversight is unevenly distributed across administrative layers. This manuscript introduces SCARA (Supply Chain Anomaly and Risk Assessment Agent), a fully integrated intelligent framework that brings together synthetic data generation, multi-category fraud simulation, geospatial logistics modelling, and a large language model (LLM)-powered analytical layer to identify and explain five canonical categories of procurement fraud: Invoice Inflation, Ghost Shipments, Product Mismatches, Product Diversion, and Shipment Delay.
The synthetic dataset was built using a statistically grounded simulation pipeline that models 50 suppliers, 20 government buyers, and 100 product types distributed across ten major Indian cities, with geospatial logistics computed using the Haversine formula. A controlled fraud injection rate of 15% introduces labelled anomalies drawn from real-world irregularities documented by Indian audit bodies. The detection layer uses Meta LLaMA 3.1 8 B Instant, served through the Groq low-latency inference API, to produce structured JSON risk reports containing an overall risk score, an executive summary, step-by-step reasoning, a hypothesized fraud category, and recommended investigative actions. An interactive Streamlit dashboard allows practitioners to explore and query the transaction data in real time. The results show that SCARA produces coherent, explainable, and practically useful fraud assessments, establishing a strong foundation for human-in-the-loop AI governance in public procurement.},
        keywords = {supply chain fraud detection; synthetic data generation; large language models; public procurement; Indian logistics; anomaly detection; explainable AI; LLaMA; Groq inference; Streamlit dashboard},
        month = {March},
        }