Sumsub Technologies

Do you know what’s behind the scene of your KYC provider?

Not all verification providers use complex technological solutions, sometimes it can be just manual checks. But how can you know what is going on? With Sumsub, your verification is not a black box. Come in, we’ll take you backstage of our decisions.

Hybrid approach provides neat results

We guarantee a 99,9% api request pass rate and it’s fixed in our service agreement. To reach such results, we provide a hybrid approach which lets wisely allocate the tasks between machines and people. This is an ultimate win-win for user experience, fraud protection and reliability.

AI Technologies

We combine the technologies developed by our R&D center with the best technologies in the market

Production data

Our database with thousands of real documents helps to train AI algorithms more efficiently

Verification experts

Our experts take care of complex cases where the special expertise is needed

Trusted technologies ensure
comprehensive checks

You get the result of verification in a couple of minutes. Explore below how many
checks we do with each profile during this time.

Catching fraudsters at the entry point

Many fraudsters can be caught at the entry point by analyzing metadata of images they upload. We dig into the characteristics of the pictures and define the possibility of fraud before identity verification.

Metadata Analysis
Metadata Analysis defines the time and place when the picture was taken and compares them to the ones when the picture was uploaded
It helps to define:
- Pictures taken in advance
- Possibly fabricated images
Device Fingerprint Analysis
Companies often struggle to identify a device when users hide the IP address or switch among browsers on the same device. Device fingerprint analysis easily identifies individual devices - this is an important step in fraud detection.
It helps to prevent:
- Multi accounting
- Identity theft
- Credit card fraud

Detecting graphic manipulation

One of the popular ways for fraudsters to cheat is using graphic editors. To define whether the picture was edited electronically, we use our unique image authenticity tool. It checks two dimensions: the device features and image structure.

Enhanced signature analysis
We use an extensive database of cameras and software signatures to determine the source of an image or detect traces of modification software with very high accuracy.
It helps to identify:
- Any traces of Photoshop and another graphics software
Pixel analysis
Pixel analysis evaluates the image authenticity in several dimensions, so we can reveal malicious manipulations.
It helps to identify:
- Areas that have been modified in the editing programs

Fraud pattern detection

Skilled fraudsters make sophisticated forgeries, buy real documents on the Internet or gain access to the documents leaked to the DarkNet. We have special mechanisms to detect such cases.

Blocklists
The system checks the documents against the Blacklist. This is the database where we bring all suspicious profiles we ever detected.
It helps to identify:
- The documents from DarkNet
- The real documents bought in the Internet
- The documents of users who were suspected in the fraud
Documents’ Face matching
When fraudsters make forgeries, they often use templates with the same faces. So, there is a chance to find several documents with the same face, but with a different name and country. With our cross-reference and duplicate search tool, we’re able to find this type of fakes.
It helps to identify:
- Fake documents with similar photos

Authenticating the document

Skilled fraudsters use sophisticated digital and manual forgeries. To fight them, we analyse protective elements to ensure the document is official. The system automatically compares stamps, fonts, colored backgrounds, holograms, watermarks, microprints and other security features to our database of 14 000+ documents.

It helps to identify:

Manual forgeries
Documents that weren’t issued by state

Payment fraud analysis

Significant group of fraudsters makes money on false chargeback requests. To stop this type of fraud, we do a set of checks using the bank card data, the data received during KYC and liveness checks.

It helps to prevent:

Illegal chargeback requests

Optical Character Recognition (OCR)

We support documents from 200+ countries and territories, and automatically check those written in Latin, Сyrillic, Semitic and Asian characters.

To verify documents we use our AI technologies: all textual data is extracted, transliterated into Latin characters and saved in the user profiles for your future reference.

MRZ reading

The system captures data from Machine Readable Zone (MRZ) of any Passport, ID, Visa and other documents

Barcode reading

The system decodes various 1D and 2D barcode formats such as PDF417, QR code, and Aztec code

Credit cards reading

The system extracts card number, cardholder’s name and expiration date from a debit/credit card

It helps to prevent:

Illegal chargeback requests

Liveness detection

Comprehensive checks are very annoying for users. What is more, gesture-based gimmicks like asking a user to blink or speak a random passcode add friction to the experience and are easily fooled by spoofing tactics.

We use our own liveness technology. It’s just one simple action for users, but a fast and secure way to ensure the presence of a user for you.

It helps to prevent attacks using:

Paper masks with eye & mouth cutouts
Hollywood masks, wax figures & lifelike dolls
Impostors, lookalikes & doppelgangers, and more

More about liveness technology

Face-based biometrics

Stolen IDs and impersonation attacks is one more possible way fraudsters choose.

Our AI technologies automatically select the best image from liveness and perform advanced facial scanning. It indicates if this image matches the photo on the ID without forcing the user to take extra action.

We use our own liveness technology. It’s just one simple action for users, but a fast and secure way to ensure the presence of a user for you.

It helps to ensure:

The person who goes through liveness is an owner of the provided documents

Fuzzy matching

When the compliance team performs AML screening, it often has to struggle with a high number of false match results. The parameters need to be wide enough to take on slight nuances in names, but not too wide to generate false positives.

To meet this challenge, we provide a fuzzy matching tool with a big number of features including: — different spelling of names (e.g. ‘Jon’ instead of ‘John’) — shortened names (e.g. ‘Elizabeth’ matches with ‘Elisa’, ‘Elsa’, ‘Beth’) — abbreviations (e.g. ‘Ltd’ instead of ‘Limited’), and many more It helps to identify: Non-exact matches during AML screening

— different spelling of names (e.g. ‘Jon’ instead of ‘John’)
— shortened names (e.g. ‘Elizabeth’ matches with ‘Elisa’, ‘Elsa’, ‘Beth’)
— abbreviations (e.g. ‘Ltd’ instead of ‘Limited’), and many more

It helps to identify:

Non-exact matches during AML screening

Customer Lifecycle management

While using Sumsub, you have a single user profile, so you get the user's entire history presented right in front of your eyes. You can see all the checks made in the past:

— what documents have been uploaded
— what checks have been passed and what are their results
— whether the user passed the questionnaire
— who from the team initiated the changes, and so on

It helps to:

Store all information about user in one place
Reduces errors during manual review

Natively integrated with all the tech you’re already using

We don’t want to change the way you work – we just want to improve it.

Would like to discuss the details with our experts?

Join 2500+ businesses
that use Sumsub