OCR Studio launches neural network for document collages, updates deepfake detection

OCR Studio has launched a neural network technology that reliably detects document collages that fraudsters use during KYC and customer onboarding. A news post from the company says the system “identifies local visual anomalies by analyzing whether a specific region differs from its surrounding image structure,” making it possible to expose high-quality document collages in which only individual characters have been replaced.

Pieced together from separate fragments of different IDs, document collages can leverage pieces of genuine IDs to bypass many verification systems without triggering standard document checks. Fraudsters insert individual characters, data fields, portraits, MRZ fragments, or other visual elements into a document image, changing birth dates, replacing portraits for account takeover, altering expiry dates during KYC, or combining a genuine document background with stolen personal data to pass digital onboarding.

“Document fraud is becoming increasingly diverse, and attack scenarios often differ significantly from one another,” says Konstantin Bulatov, chief technology officer of OCR Studio. “Even advanced anti-fraud systems that reliably detect common forgery types can still fail when they face a manipulation pattern they were not specifically trained to recognize.”

OCR Studio’s neural network takes in the big picture, analyzing both individual fields and the overall visual consistency of the document, flagging areas that don’t match the surrounding content but may be invisible to the human eye. The technology’s architecture shrinks computer‑vision models by more than 40 times while maintaining accuracy. It has been integrated into OCR Studio’s flagship document forensics system, OCR ID-verify, which is delivered as an on-premise tool and does not store or transmit ID data to external services, helping comply with data privacy frameworks such as GDPR, PIPL, and CCPA.

Deepfake detection tool debuting at GITEX AI Europe

OCR Studios has also launched new deepfake detection technology. According to the company’s website, the system exposes AI-generated and morphed identity documents created with AI models including ChatGPT, Nano Banana, Grok, Midjourney and others.

The new system analyzes the underlying structure of an ID image and detects low-level artifacts left by generation or editing algorithms. It does not rely on document context or search for logical inconsistencies, making it resilient to high-quality deepfakes that contain no visible artifacts. Integrated into OCR ID-verify, the product is debuting at GITEX AI Europe, which runs from June 30 to July 1, 2026 at Messe Berlin.

Article Topics

AI fraud  |  deepfake detection  |  generative AI  |  identity document  |  morphing attack  |  OCR Studio