Lightweight Embedded Framework for Edge Document Authenticity Detection

Salma Samiei

Keywords: Document Forgery Detection, Document Liveness Detection, OCR and Textual Analysis, Portrait Consistency, Hologram Detection

TL;DR: Lightweight Multi-Modal Pipeline for Identity Document Verification on Resource-Constrained Devices

Abstract: The rise of generative models has intensified the threat of highly realistic identity-document forgeries. We propose a compact, system-level framework that unifies visual, textual, and security cues within a calibrated decision pipeline. Unlike isolated detectors, our approach integrates heterogeneous modules into a resource-constrained architecture, ensuring both robustness and interpretability. Evaluated on public benchmarks (MIDV, DLC, FMIDV, MIDV-Holo) and a large in-house dataset spanning 128 countries, the framework achieves competitive performance in both detection and localization. Its efficiency and accuracy make it a practical candidate for mobile and embedded ID verification.

Submission Number: 4