A Geo-Tagged AI and Blockchain Framework for Transparent NGO Donation Tracking with Fraud Detection and Real-Time Impact Visualization

Abstract

Donors give to non-governmental organisations on trust, and when that trust is broken — by fund misuse, phantom beneficiaries, or fabricated field reports — giving collapses, hurting the communities NGOs serve. This paper presents a donation-tracking framework that earns trust through evidence rather than assertion, integrating three technologies that each cover a different failure mode: geotagging binds every disbursement to a verified location, an AI model flags fraudulent donations, and a blockchain ledger makes the record tamper-evident. The central, measured finding is that these integrate rather than merely coexist: a Random-Forest fraud detector trained on twelve donation features reaches a 0.924 F1, and adding the geo-tag features lifts its recall from 0.69 to 0.90 and F1 from 0.799 to 0.924, with the donation-to-NGO GPS distance the single most important feature — geo-tagging is not a separate dashboard but a decisive fraud signal. A geo-verification layer rejects 100 percent of GPSspoofing, metadata-tampering, and duplicate-proof attacks on 2,000 trials. The blockchain layer anchors a donation for about USD 0.003 when batched and verifies it off-chain in 0.016 ms, scaling to one hundred thousand donations at flat latency. With a five-layer architecture, a documented dataset, statistical validation, a threat model, and a privacy design that keeps personal data off-chain, the framework turns transparent giving from a slogan into a validated system.