Digital Identity Verification Using Machine Learning To Reduce Fraud In Micro-Lending And Enhance Credit Risk Assessment

Authors: Dr. Pankaj Malik, Kanishka Raghuwanshi, Moksha Jain, Manmohan Rajput, Mohd. Shayaan Dehlvi

Abstract: Micro-lending institutions play a vital role in promoting financial inclusion, but they are highly vulnerable to identity fraud, impersonation, and inaccurate credit risk assessment due to limited borrower histories. Traditional Know Your Customer (KYC) and credit scoring approaches are often manual, time-consuming, and ineffective against sophisticated fraud techniques such as synthetic identities. This study proposes an integrated machine learning–based digital identity verification framework to reduce fraud in micro-lending and enhance credit risk modeling. The proposed system combines document verification using optical character recognition, biometric face matching with liveness detection, and device–behavioral analytics to generate an identity confidence score. This score is then incorporated into advanced credit risk models to improve default prediction accuracy. Experimental evaluation conducted on a micro-lending dataset demonstrates that the proposed identity verification module achieves a fraud detection accuracy of 94.6%, with a precision of 92.8% and recall of 91.3%. When integrated into credit risk models, the enhanced framework improves the ROC-AUC from 0.74 to 0.86, and reduces false loan approvals by 31% compared to conventional models without identity features. These results confirm that ML-driven digital identity verification significantly strengthens fraud prevention mechanisms and improves credit risk assessment, enabling secure and scalable micro-lending operations while supporting broader financial inclusion.