24.06.2025

Smart Engines Unveils a Multimodal AI System for Document Forgery Detection

Smart Engines has introduced the first multimodal AI technology designed to detect forged documents. The solution verifies the authenticity of documents from around the world for using 600 parameters and identifies complex forgeries during both in-person and remote verification, including KYC/AML processes. The model is built on the principles of explainable AI.

Multimodal AI belongs to the most advanced class of neural network architectures, capable of simultaneously processing different types of input data—text, images, audio, and video. Smart Engines researchers have applied this capability to the field of multimodal forensics. The new system can simultaneously process document images in optical, ultraviolet, and infrared spectra; video sequences featuring the document; text fields; NFC chip data; barcodes; metadata; and signatures. This enables highly accurate ID scanning and detection of forged or altered documents.

The model analyzes each page and element of a document individually while also checking consistency across all components. This approach makes it possible to reveal hidden discrepancies, including incorrect page numbering, differences in microprinting, or mismatches between NFC chip data and the information in the VIZ and MRZ.

This integrated method allows the system to detect complex forgeries in which individual pages may be replaced, a photo may be reattached, the order of security elements may be deepfake-based forgeries.

The technology operates entirely within the customer’s IT infrastructure and does not require a graphics processor. Images and document data are never transferred to third-party services, cloud verifiers, or crowdsourcing platforms.