Idea
A dataset and benchmark platform for developers and companies to improve digital ID forgery detection accuracy in KYC processes
Research Paper
Core Innovation
This paper presents FantasyID, a novel dataset that simulates realistic identity document forgeries without using real legal documents or synthetic faces. It includes diverse ID designs and languages and mimics real-world KYC scenarios with printed and captured cards. This approach exposes weaknesses in current forgery detection models by generating high false negative rates, providing a challenging benchmark for future improvements.
Market Size (TAM)
$2–10B TAM, $1–2B SAM; assumption: growing global digital identity verification market and regulatory compliance needs.
Potential Customers & Pain Points
- Financial Institutions Needing Reliable KYC Verification
- Identity Verification Service Providers Seeking Robust Detection Benchmarks
- AI Developers Lacking Realistic Forgery Datasets
Business Model
Freemium access to the dataset with premium consulting and custom dataset generation services for enterprises
Competitive Landscape
- IDnow
- Jumio
- Onfido
Implementation Challenges
- Dataset adoption by industry
- Integration with existing KYC systems
- Continuous update to cover new forgery techniques
Validation Strategy
- Release dataset to public and gather user feedback
- Benchmark leading forgery detection models on FantasyID
- Collaborate with industry partners for real-world testing
Research Paper Overview
FantasyID: A dataset for detecting digital manipulations of ID-documents
Summary
This paper introduces FantasyID, a publicly available dataset designed to detect forged identity documents. It mimics real-world IDs without using legal documents or generated faces, includes diverse designs and languages, and simulates realistic KYC scenarios with printed and captured cards. The dataset challenges current state-of-the-art forgery detection algorithms, showing high false negative rates, making it a robust benchmark for improving detection systems.