Computation Document Forensic and Liveness

Computation Document Forensic AI — a multipurpose AI system for digital and physical document forgery detection and forgery prevention in user identification and verification procedures; financial, banking and insurance services; communication services; remote public services and ticket sales for intercity and international transportation. Thanks to the latest AI developments, the system detects and prevents various document forgery attacks. The special Liveness block reflects document presentation attacks with a digital copy instead of the original. The Green AI approach ensures the highest speed and energy efficiency.

Computation Document Forensic AI is an integral part of the document recognition system and implements the principle of fraud detection via recognition, in which AI uses all the data and signs detected at all document recognition stages. The system allows you to detect image forgery in scanned documents, photos, videos, and multispectral images captured with the help of special devices. Computing Document Forensic AI uses several specialized AI models for anomalies and fraud detection.

In the passport recognition process, Computing Document Forensic AI allows you to detect document forgery attempts in the optical and infrared ranges, as well as when documents are illuminated with UV light with a wavelength of 365 nm. For the passport recognition process on smartphones, there was designed the proprietary video stream analysis system for repelling document presentation attacks with a “poisoned” (physically modified) document image instead of the original.

AI is completely autonomous and works automatically, all calculations are performed in the device’s memory and are NOT transferred. AI does NOT require a network connection and does NOT store any data. Thus, it prevents the disclosure of personal and biometric data from the recognition system. The system complies with GDPR and other local and international legislative acts in the field of personal data protection.

Computation Document Forensic AI includes the following models:

• Image Fraud AI — AI that analyzes the image container and detects anomalies and discrepancies related to fraudulent activities.
• Pixel Fraud AI — a special AI model that detects synthesis or interference in the document images.
• Font Fraud AI — a unique deep neural network (DNN) that provides font verification in the documents.
• Text Fraud AI — a special AI model that detects physical and digital anomalies and discrepancies in the method and parameters of text document data.
• Template Fraud AI — an AI system, that detects anomalies and discrepancies in a document template, as well as verifies the authenticity and the security elements of a document template.
• Photo Fraud AI — an AI system that detects physical and digital anomalies of a facial photo on the document.
• Data Fraud AI — an AI model that provides fraud detection and document data verification. It is designed to detect data discrepancies and repel attacks on recognition.
• Stamp&Seal Fraud AI — an AI model that provides search, recognition, verification, and identification of forged seals and stamps on documents.
• Liveness AI — a family of AI models designed to repel document presentation attacks with a digital or physical copy.
• Holo AI — a proprietary model of search, localization and identification of security features such as holograms and other OVDs (Optically Variable Device), based on a combination of algorithmic and neural network temporal analysis of the document video stream.
• Multispectral Fraud AI — search for physical and digital document forgery thanks to intelligent analysis of multispectral document images.

Send Request