Automating the identity verification process through Artificial Intelligence

As much as gaining customer's confidence and trust is imperative, it is equally important for businesses to verify the information that their customer's provide. Businesses seek to adopt to Know Your Customer (KYC)/ Business (KYB), Anti-Money Laundering (AML) rules, Customer Due Diligence (CDD) checks and similar frameworks built around verifying the identity information of their clients such as name, address, photograph, original documents, etc in order to prevent identity theft. 

Now, typical institutions in India and other countries spend an immoderate amount of time onboarding a client with dependence on manual interventions which turn out to be expensive, laborious, inaccurate and, inefficient. Any lapses in controls have resulted in regulators tightening supervision on the industry as a whole. 

Technology's impact on modern day has been stupendous. As the world holistically is embracing a fast-paced environment, businesses are keen on adopting automation through advanced technologies such as Artificial Intelligence, Machine Learning, etc.

Artificial Intelligence is a collection of related technologies which has the potential to automate workflows, quickly analyse large volumes of data of different types and has contributed to bountiful innovations in recent times. 

The Machine learning knowledge used by AI in its hybrid systems assist in tracking the malicious transactions and ultimately help in combating risks of fraud and improves the operational efficiency. 

AI not only saves money but also efforts and valuable time of people who are otherwise required to put tremendous amounts of manual effort. It helps streamline the regulation and compliance process, eliminate fraud, solves for human errors, automates repeatable tasks, and helps reduce time and cost. 

Illustrated below are a few preeminent ways in which AI can help improve KYC and client onboarding processes:

A. Facial Biometric Recognition:

The face-match service is the most efficient way to check the customer's facial identity while performing online transactions. These help verify the legitimacy of the customer by authenticating the photo ID submitted using unique mathematical and dynamic patterns. 

The face recognition can be framed in a three-step process as described below:

• Step 1- Source Image: Captures facial features from ID cards or other identity documents

• Step 2- Target Image: Prompts the user to take a photograph of themselves using selfie face recognition technology for biometric liveness detection purposes

• Step 3- Match Score: A match score for the target and source image is provided in real-time, with a yes or no result for validation purposes. Businesses can set a predefined match score for auto-validation, any score below the limit will prompt re-initiation of the process which could be either carried out through a live selfie KYC or Video KYC.

Application:

This simple face match process provides a robust identity establishment through facial biometric verification for businesses across a spectrum of industries. A few examples include validating the identity of the person collecting the cash during a lending disbursement process with the KYC identity. It is also widely used in the process of account opening for Bank/ Demat, etc where the photo in proof of identity document is matched with the person who is performing the KYC process.

B. Image classification, parsing and verification: 

The image processing technique is a method to extract information from an input image and further analyse the information through cognitive analysis and pattern recognition. It caters real-time scanning of proof documents, performs discrepancy checks and provides instant validation scores that can be used to automate the decision-making process.

Key phases of image processing include:

• Step 1- Image Acquisition: is the process of capturing an image and converting it into a digital image file.

• Step 2- Image Enhancement: technique to improve the quality of an image in order to extract hidden information

• Step 3- Image restoration: is also a process used to enhance the quality of an image. This process is based on probabilistic and mathematical frameworks and can be used to get rid of blur, noise, missing pixels, camera misfocus, watermarks, and other impurities that may hinder the image quality. Image restoration can also include colour processing, compression/ decompression of an image which relates to changing the size and resolution of an image.

• Step 4- Morphological Processing: describes the shapes and structures of the objects in an image. Morphological processing techniques can help in building data sets for evaluation in the Machine learning process.

• Step 5- Image Recognition: is the process of identifying specific features of objects placed in an image.

• Step 6- Representation and Description: the data processed during the image analysis are generally raw output in the form of an array of numbers and values. Visualisation tools are deployed to represent these values in the required format.

          

Application:

Helps identify the type of the identity proof uploaded (Eg: PAN card, Driver’s license, Passport, etc) and identifies for any tampering in the document by evaluating fonts, image pixel and edits/ manipulation of the document to check for digital forgery of the displayed ID proof.

Handwritten signature verification on a blank paper is another prominent step in the KYC process for many financial institutions and regulated entities in India. With the help of  image processing and classification enabled by Artificial Neural Networks, Digio helps businesses automate the signature verification process. Provided below is a brief of the process: 

C. Optical Character Recognition (OCR):

OCR is a technology that enables you to convert different types of documents, such as scanned paper, PDF files or images captured by a digital camera into an editable and searchable data. In real life scenarios, image processing and OCR mostly operate together.

The generic process for performing the character recognition is as follows:

• Step- 1: Scanned Identity document is sent to Digio for processing

• Step- 2: The file is processed and converted into a raw text file, which is a computer-readable version of the identity document

• Step-3: Machine learning models are used to interpret the data and send back information in the document in a JSON format.

            

Sample PAN OCR from Digio’s Dashboard.

Application:

OCR can be carried out on identity documents, proofs, agreements, restaurant bills, academic certificates, Income proofs, etc. The identity and document verification is applicable across industries and use cases and Digio’s image processing technology has successfully helped businesses save substantial amounts of time.

Additional solutions which can be implemented through the help of Artificial intelligence

D. Unstructured Data Analysis: 

Artificial Intelligence's NLP and defined Machine learning algorithms enhance the client onboarding process through intelligent document scanning by reading vast amounts of unstructured data in any language including extracting metadata, identifying and deciphering the intent or purpose of specific parts of a document. Processing the customer data involves a series of techniques such as reducing noise, eliminating irrelevant information and bifurcating data into synchronised content for decision-making. 

Application:

The unstructured data analysis combined with OCR can help validate contracts and agreements and aid in database search capabilities including comparing / editing extracted data to build touch points for fraud and compliance detection. It caters to the automation of several processes including internal contract management, vendor life-cycle management, etc. The language analysis used by companies to identify and monitor any form of censored content in communications and help them manage corporate governance policies. 

E. Link analysis: 

Link or network analysis focuses on the relationships between entities rather than attributes of entities. AI analytical frameworks can be used to anatomize large numbers of objects of different categories and help link relationships between them to comprehend evidence which could not have been gathers from a single piece of information. 

Application:

Link analysis comes handy for financials institutions in their decision-making processes especially with respect to evaluating the eligibility of a user to participate in the lending or investment process. It helps in preventing money laundering by tracing fraudulent transactions and by analysing the relationships between transactions and predicting patterns. Link Analysis can also be used to automate user face recognition by matching the unique biometric identifier of the individual acquired during a live KYC through selfie or video with the existing client database. 

Every organization should develop a risk-based approach in their client onboarding journey by considering multiple factors such as Industry, regulations, clientele, demography, etc. Interested in knowing how best these technologies can be put to use in your business? Reach out to our experts at bd@digio.in