A primer on using AI for ID document verification

Today’s AI-driven KYC platforms help banks, fintechs, insurers and more onboard customers and catch fraud at a higher rate—with less dropoff—than ever before. Through the sequential use of proprietary AI algorithms, these platforms maximize accuracy while trimming minutes off ID processing times—without sacrificing security.

Advanced image processing 

The central part of KYC is, of course, identity verification. As part of the onboarding flow, users typically submit images of their identity documents for rigorous assessment and risk-scoring. But before drawing any conclusions, AI models must detect, recognize, and interpret the photographic input.

To build an effective and scalable solution, an AI-driven KYC provider creates a set of algorithms that can treat thousands of different document types in a consistent and standardized way. Across document types, certain foundational steps remain the same: models require front-facing and angled images of a user’s document, alongside manual input of the document information.

While the typical onboarding flow guides users through taking these pictures, the platform must account for the extreme degrees of domain variance that are inherent to this type of data collection. A chain of algorithms detects, adjusts, classifies, and reads each document image, in spite of variances in angles, shadows, lighting, resolution, etc. Then, based on the data, security features, and consistency checks, the platform draws conclusions that facilitate authentication of a user’s submission.

Document detection and recognition

When the user submits a front-facing image of their ID, the Document Detection algorithm identifies the document in the image. From there, the algorithm generates coordinates of the borders of the document, which connect to form what is known as a “bounding box.” The technology uses these coordinates to crop and zoom in on the document in question, preserving all areas of the document that would have useful information or security features.

The cropped image generated by the Document Detector is used to identify the model of the document (origin country, document type). Once it has been recognized, the platform can start processing and assessing the information within the image.

Visual Inspection Zone detection and reading

The next task is to collect and analyze the information within the document that was just recognized. Every document has its own specifications and security features, starting with where information is positioned on the document.

Once a document has been cropped and recognized, the Visual Inspection Zone (VIZ) Detector will map coordinates upon the image that dictate where to find the information needed to assess the validity of the document.

Naturally, the locations of different information fields vary by document. For instance, a French ID card will display a person’s birthday in the middle-right area of the card’s front-side, whereas an American passport will show the same information on the middle-left.

With the VIZ fields mapped, the algorithm can crop and zoom in on each field to extract the information needed. The same mechanism of “bounding boxes” used in the Document Detector is applied to each VIZ field in the document.

Then, the VIZ Reader is tasked with reading these fields. Once the information has been read, conclusions about the user’s identity and personal details can be inferred. For instance, one of the basic checks verifies whether a user is of age to be accepted. This requires detecting where the birth date is on a document, reading the date, comparing this with today’s date, and finally calculating whether the person is the age required.

 Another application of the VIZ reading process is the comparison of a user’s manually input data to the automated reading results. This not only weeds out inconsistent data entry, but also flags errors the algorithm may have made—educating the Optical Character Recognition (OCR) algorithms to run better in the future.

Delivering Better User Experiences

Ultimately, carefully designed and executed AI technology serves to process identifications more accurately, at the fastest speed possible. Through the application of this technology, companies can reduce their total end-to-end KYC process time to just a minute or two, with a very high degree of fraud-detection accuracy. As users demand ever more seamless onboarding processes, reducing these processing times becomes critical to preventing dropoff and delivering a high-quality customer experience.