Step 1
Image quality control
The Web SDK shows a frame and prompts, and the document image quality check filters out blurred, dark and cropped shots. A bad shot should not turn into a false suspicion.
Document authenticity check
Automatic authenticity check of a document from a photo or scan: NeuroVision looks for traces of editing, a pasted-in photograph, screen capture, printouts and signs of AI generation, verifies the MRZ and check digits and, for a Russian passport, also checks it against the Ministry of Internal Affairs (MVD) database. The result arrives in the API together with the recognised data.
Document check Specimen · ID document
- Traces of editing not found isEdited
- Screen capture not found isDisplay
- Printout or copy not found isPrinted
- Signs of AI generation not found anti-fraud
- MRZ and check digits match MRZ
- Result passed No signs of forgery found
- 40+ document anti-fraud checks
- 16,000+ document types from 195+ countries and territories
- Russian passport against the MVD database — 5 ₽
Data is stored in Russia Russian Software Registry No. 19972 Compliance with 152-FZ
What an online document authenticity check is
Validation is not authenticity
The MRZ algorithm is published in ICAO Doc 9303 — check digits can be recalculated by whoever forged the document. That is why NeuroVision analyses both the data and the image
- 01
Authenticity — has it been forged
Are there traces of a graphics editor, a pasted-in photograph, screen capture or signs of neural-network generation. Answered by document anti-fraud.
- 02
Validity — has it lost force
A passport may be genuine but replaced, lost or expired. Answered by the register of the issuing authority; for a Russian passport, the MVD database.
- 03
Data correctness — validation
Number format, date logic, MRZ check digits, MRZ matching the visual zone. A necessary condition, but not a sufficient one.
- 04
Why the naked eye is not enough with a photo
The employee sees a screen, not a document: a hologram cannot be tilted and neat editing goes unnoticed. Forgeries have become cheaper, and with hundreds of applications a day manual checking is a bottleneck.
- 3 questions to any document
- 40+ image anti-fraud checks
- 1 request — data and checks
Types of forged documents
Detecting forged documents starts with understanding how they are made. Six typical schemes — and what NeuroVision finds in each. The illustrations show a specimen Russian passport with fictitious data. A sign found is a reason for a manual check, not an accusation against the client: glare, an old form or a poor camera also produce a strange picture.
A document assembled from a template
A fully fabricated document: the background, fields and photograph are assembled in a graphics editor from a ready-made template. Such templates are sold online and used dozens of times.
What NeuroVision finds: Deviation from the reference for the document type: field layout, fonts, mandatory design elements.
Altered data
Partial forgery: the document is genuine, but the surname, date of birth or number has been replaced — to get round an age restriction or to pass off someone else's document as one's own.
What NeuroVision finds: Traces of editing in the field area, a different font and spacing, a mismatch between the field and the MRZ. A passport forgery is also exposed by check digits if the MRZ was recalculated with an error.
A swapped photograph
Your own photo is pasted into someone else's document — physically, on the form, or digitally, on the image.
What NeuroVision finds: Uneven edges and shadows around the photo, a break in the security background, traces of a layer overlay.
A document generated by a neural network
The whole document image is created by a generative model: a plausible background, the face of a non-existent person, data without typos. Such images look cleaner than real photographs.
What NeuroVision finds: Signs of a synthetic image: unnaturally even light and noise, artefacts in fine patterns and in characters.
Screen photo
The client photographs not the document but its image on a monitor or phone: someone else's photo from a chat, a leak or an edited file.
What NeuroVision finds: Moiré, a pixel grid, glare and the screen frame. In the API response — the check isDisplay.
Printout and photocopy
Instead of the original, a colour printout or photocopy is shot — often of an already edited document.
What NeuroVision finds: Paper texture, print halftone, loss of colour and flat “security” elements. In the API response — the check isPrinted.
What signs of forgery NeuroVision looks for
Document anti-fraud is 40+ checks grouped into six areas. They run together with recognition in a single request, and the check for forged documents requires no extra action from the client. Technical details — in the article on traces of editing.
-
Image integrity
Document integrity check: local edits, cloning and erasing of areas, splices, traces of re-saving, redrawn digits in the details.
-
Medium: screen, printout, copy
We tell a live capture of a document from a photo of a screen, a colour printout or a photocopy.
-
Layout and design
Comparison with the reference for the document type: field layout, fonts, mandatory visible design elements, a signature and stamp where they are provided for.
-
Data and MRZ
Check digits, MRZ matching the visual zone, date logic, validity period, issuing-unit code and issuing authority for a Russian passport. More about the MRZ.
-
Holder's photograph
Signs of a pasted or replaced photo: edges, shadows, a break in the security background around the picture.
-
AI generation and document condition
Signs of an image created by a neural network; physical tampering and wear of the document that make the data impossible to read reliably.
How the automatic document authenticity check works
The authenticity check runs at the same time as recognition: the client takes one shot, and the system returns the data, the checks and the result in a single request. More about recognition — Recognition of Russian internal and international passports.
Step 2
Document type detection
The model recognises the document among 16,000+ types from 195+ countries and territories and selects the reference against which the layout will be compared.
Step 3
Image analysis
We look for traces of editing, a pasted-in photo, screen capture, a printout and signs of generation. Each check gets its own status.
Step 4
Data and MRZ check
We recognise the fields, recalculate the check digits, and compare the MRZ with the visual zone and the date logic. For a Russian passport, a validity check against the MVD database can be run on request.
Step 5
Result and decision
Check statuses, recognised fields and the result arrive in JSON. The decision — approve, reject or send for manual review — is made by you under your own rules.
Checking document security features: what is visible in a photo
- Traces of digital editing — almost invisible to the eye
- Screen capture, printout
- MRZ and check digits
- The holder's photograph and its edges
- Fonts and field layout
- Security background and guilloche grid — if the shot is sharp
- Ultraviolet elements: glow of threads, fibres and inks
- Infrared inks
- Watermark
- Laser perforation of the number
- Holograms and optically variable elements — a photo shows glare, but not authenticity
- Microprinting — as a rule, not visible
- The security features of passports, ID cards and driving licences are designed for work with the original: they are examined under ultraviolet light, against the light or at an angle. An ordinary phone photo does not create such conditions.
- What “no signs of forgery found” means. The image has not been altered, a real document was photographed, and its data are consistent. This is not a conclusion about the physical authenticity of the form: if the procedure requires checking the original — a UV lamp, light transmission, expert examination — the online check does not replace it, but weeds out forgeries earlier.
- Which security features a Russian passport has and which of them are visible in a photo — in the article “Security features of a Russian passport: microprinting, MRZ and signs of forgery”.
Online check, inspection of the original and expert examination
The three ways of checking a document answer different questions. An online check does not replace expert examination, but it solves a task the other two do not: checking thousands of documents that clients send as photographs. A conclusion on the authenticity of the form that carries evidential weight is given only by an expert examination.
-
Inspection of the original
An employee examines the form presented to them: with a lamp and against the light.
Who and where
- Carried out by — the employee to whom the document is presented
- Where — in the office
What is checked
- What is analysed — the physical form in hand
- UV, watermark, perforation — yes, with a lamp and against the light
- Traces of digital editing — no, they are not on the form
Speed and result
- Speed — minutes
- Output — “matches the specimen” or “doubts remain”
-
NeuroVision online check
The system checks the document image and its data from a photo or scan.
Who and where
- Carried out by — the system, from a photo or scan
- Where — in an app, on a website, in the office
What is checked
- What is analysed — the document image and its data
- UV, watermark, perforation — no
- Traces of digital editing — yes
Speed and result
- Speed — automatic, in a single request
- Output — check statuses and a result: “no signs of forgery found” or “signs found”
-
Document expert examination
A forensic expert examines the document and the way it was made.
Who and where
- Carried out by — a forensic expert
- Where — in an expert unit
What is checked
- What is analysed — the document and the way it was made
- UV, watermark, perforation — yes
- Traces of digital editing — yes, if a file is examined
Speed and result
- Speed — days
- Output — an expert's conclusion
Authenticity and validity: a passport against the MVD database
Anti-fraud answers whether the image has been forged. Validity is a separate question: a genuine passport may have been replaced, lost or expired, and the number of a valid passport can be entered on a forged form. The Russian passport validity check queries the MVD database via SMEV by series, number, issue date, full name and date of birth and costs 5 ₽. Together the two checks close both questions. Catalogue of database checks
Authenticity check API response
One request — recognised fields, anti-fraud check statuses and the document type. Passport validity against the MVD database is started by a separate method and arrives asynchronously.
- Fields in JSON
- Confidence for every field
- Registration and MRZ
- Sandbox and documentation
Terms of connection
A free one-month test, a dedicated manager, 24/7 support, payment only for recognised documents.
Who needs a document forgery check
A document forgery check is needed wherever documents are sent as photographs and a mistake costs money. The whole anti-fraud platform — on the anti-fraud page.
-
Banks
- Remote account opening and card issuance
- Cut off an edited passport before the decision Learn more
-
Microfinance and fintech
- Loans against someone else's and altered documents
- A decision without an operator Learn more
-
Pawnshops
- The pledgor's document in an online valuation and redemption
- Printouts and screen photos Learn more
-
Marketplaces and classifieds
- The seller's document for a trust badge or a contract
- Reused templates Learn more
-
Car sharing and rental
- Driving licence and passport at registration
- Altered date of birth and experience
-
HR and personnel departments
- Documents of candidates and foreign employees
- Remote onboarding
How much a document authenticity check costs
There is no subscription fee: you pay only for the checks performed. Prices are the same as on the anti-fraud page and in the catalogue of checks.
Calculate the cost- 01
From 15 ₽ per document, each next one in the same session — from +8 ₽
- 02
Russian passport validity against the MVD database — 5 ₽
- 03
First month — a free test on your own documents
Glossary of document authenticity checks
Short definitions of terms found in the API response and in the requirements for customer checks.
- Document authenticity
- The document was produced by an authorised body and has not been altered. A photo allows checking the authenticity of the image and data; the physical authenticity of the form is established by expert examination.
- Document validity
- The document has not lost force: it has not been replaced, lost or expired. Checked against the register of the issuing authority.
- Document validation
- Checking the correctness of data and format: number length, date logic, check digits, consistency of fields.
- Visual inspection zone (VIZ)
- The part of the page where data is printed for human reading: full name, dates, number.
- MRZ
- The machine-readable zone — the lines at the bottom of the page with data and check digits under ICAO Doc 9303. More about the MRZ.
- Check digit
- A digit calculated by an algorithm from the other MRZ characters. It catches reading errors and crude edits, but does not prove authenticity.
- Optically variable element
- A hologram or ink that changes colour and pattern when tilted. It cannot be checked on a static photo.
- Guilloche grid
- A complex security pattern of interwoven lines. A break in the pattern is a common trace of editing or a pasted photo.
- Microprinting
- Text so small that it is visible only under magnification. Whether a Russian passport has it — in our article.
- Screen photo
- A picture of the document image on a monitor or phone instead of the document itself. In the API — the check
isDisplay.
Why NeuroVision is trusted with document checks
The authenticity check is part of the customer verification platform: the same models, infrastructure and security requirements.
NeuroVision — an expert in computer vision and AI
AI-OCR works as part of customer verification (KYC), trusted by the largest companies in fintech, e-commerce and transport services.
We develop identification algorithms that automate checks, detect fraud and help meet regulatory requirements.
A Russian developer
Skolkovo resident, software in the Russian Software Registry (No. 19972), Rospatent certificate No. 2023661508.
Check your clients' documents for free
Send us your task — we will give you test access to the API and SDK for a month and show which checks work on your documents.
Can't put your request into words? Book a call and our team will contact you at a convenient time
Book a call
The company is included in the register of IT companies and the register of personal data operators of the Russian Ministry of Digital Development.
Address: Moscow, Presnenskaya Embankment 12 — Moscow City business centre, Federation Tower. Nearest metro stations: Delovoy Tsentr and Mezhdunarodnaya.
Read also
How forgeries are made and how they are found — in our blog.
408 22
Anti-Fraud, Liveness and Deepfake Protection
Anti-Fraud in Document Verification: How to Detect Forgeries and Traces of Editing
413 13
KYC Document Verification
How KYC Systems Detect Forged Documents
551 13
KYC Document Verification
Passport machine-readable zone verification: how the MRZ helps detect forgeries in KYC
80 29
KYC, AML and Digital Identification
115-FZ: What Businesses Are Required to Do and How to Meet the Law’s Requirements
Frequently asked questions about document authenticity checks
Answers about passport types, the registration page, accuracy, data storage and integration
Upload a photo or scan to a service that analyses the document image and data. In a single request NeuroVision looks for traces of editing, a pasted-in photo, screen capture, a printout and signs of generation, and verifies the MRZ and check digits. For a Russian passport, a validity check against the MVD database can be added.
No. UV and IR elements, a watermark and laser perforation are visible only with a lamp or against the light, and an ordinary phone camera does not create such conditions. A photo allows checking the document image and data; if the procedure requires inspecting the original, an online check does not replace it.
Authenticity — whether the document has been forged; validity — whether it has lost force. A genuine passport may be invalid if it was replaced or reported lost. That is why it is worth running document anti-fraud and the MVD database check together.
It is a check of the correctness of data and format: number length, date logic, MRZ check digits, consistency of the fields with each other. Validation catches errors and crude forgeries, but a neatly edited document with a recalculated MRZ will pass it — such a document is found by image analysis.
Yes, signs of AI generation are part of the 40+ anti-fraud checks. The system looks for traces of a synthetic image: unnaturally even light and noise, artefacts in fine patterns and characters.
By the signs of the medium: moiré, a pixel grid and screen glare, paper texture and print halftone. In the API response these are the isDisplay and isPrinted checks.
Not necessarily. The MRZ algorithm is published, and the check digits can be recalculated by whoever forged the document. A match of the digits confirms only the integrity of the record, so a passport forgery is exposed together with image analysis.
No. An online check shows whether there are signs of forgery on the image and helps decide whom to let through and whom to check manually. A conclusion on the authenticity of the form that carries evidential weight is given only by an expert examination.
Passports, ID cards, driving licences, residence permits and other documents — 16,000+ types from 195+ countries and territories. The set of checks depends on the document type: for example, not every form has an MRZ.
Do not draw a conclusion from a single sign: ask the client to retake the document in good light or send the application for manual review. The decision is made by your company under its own procedure; the API response shows exactly which check was triggered.
The anti-fraud check — from 15 ₽ per document, each next document in the same session — from +8 ₽. Passport validity against the MVD database — 5 ₽. The first month is a free test.
In Russia: in the NeuroVision cloud or on your servers when installed in your perimeter. Processing complies with 152-FZ; with an on-premises installation, the images do not leave your infrastructure.
