Font Size: A A A

Interviews

Automation against document fraud

Martin Rehak, Founder and CEO of Resistant AI, discusses how next generation automation can combat document fraud.

Technology, and indeed the pandemic, has totally changed the way we interact. Without face to face interaction, establishing trust in an identity has become more of a challenge. In markets such as financial services it is vital to ensure that people are who they say they are and the verification documents they provide are genuine.

Juniper Research predicated that e-commerce losses to payment fraud would exceed $48 billion by 2023. So, when automation has replaced manual tasks and the customer experience is expected to be streamlined and hassle free, how do you remove the threats of fraud?

The cost of fraud

Any organisation whose services are driven by proof of identity or Know Your Customer (KYC) open themselves up to manipulation, be it for loans, mortgages, bank accounts and so on. However, when competition is high and the need to on-board customers seamlessly is essential, often the lines get blurred.

Document fraud falls into four categories:
1. Image fraud: the ID document is provided as an image which is easier to manipulate.
2. Illegitimate documents: Completely false documents that have characteristics missing.
3. False documents: Use of ID which belongs to another person.
4. Modified documents: An original document that is altered.

Recent data gathered by Resistant AI has revealed, for example, that 17 per cent of bank statements received as part of KYC in support of lending applications are tampered with. In addition, Know Your User (KYU) processes, which include merchants, fintechs and the B2B ecosystem, among others, are also subject to fraud. Furthermore, 11pc of UK payslips submitted as part of digital loan applications are altered or forged. This shows that highly automated workflows used by financial services are particularly vulnerable to sophisticated manipulation by fraudsters.

The consequences are apparent in the 5pc of loans that are underwritten every year against forged documents where every dollar lost incurs a further $3 in fees, labour costs and recovery expenses; in the 6pc of global healthcare spending lost annually to fraud – around $500 billion in 2020; and in the 3pc to 4pc of fraudulent insurance claims costing carriers somewhere between 5pc and 10pc of their annual revenues.

Know who to trust

National identity cards and drivers licences provide proof of who we are; payslips and bank statements, proof of what we earn; and utility bills, proof of where we live. Whilst manipulated documentation will continue to be used in fraudulent applications, it is possible to protect highly automated workflows from document forgery.

Visual and structural modelling can be used to assess the look and feel of specific types of documents provided by third parties, comparing them against examples of authentic documents provided by document originators – banks, utility companies, government agencies. There are also ways of detecting manipulation of documents with graphics editors, which delivers a success rate of 75-80%.

Enter the big guns

Any modification to a document generally indicates fraud, so the key is for that to be identified as soon as possible during the customer onboarding journey. It is better for businesses to be able to know instantly which documents to trust, rather than analysts spending time manually trying to detect anomalies that are impossible to spot with the human eye.

Instead, Artificial Intelligence (AI) and Machine Learning (ML) takes on a forensic analysis role that literally examines a document pixel by pixel, never relying on a single indicator to make a decision. Multiple indicators, detectors and parameters are employed to show up the minutest anomalies and this data is built up to establish patterns and develop modelling on a constant basis. Once documents such as payslips or bank statements are classified, they are cross-referenced against legitimate templates to identify any mismatches. This way, the analytical engine continuously learns as the volume of documents processed increases.

For digital-first businesses this can deliver significant benefits including:

• Faster customer approvals
• Risk teams can focus on real document fraud cases
• Workflows can be safely automated
• Fraud risk gets lowered even with high-risk applicants
• Models can be tailored for use cases to increase precision

The holy grail is to identify a customer and their related documents as being legitimate as soon as possible. That way they can be provided with their goods or services promptly and the entire process is transparent and seamless. Striking the right balance between accessibility and security is vital – fraud risk should be minimised, customer journey should be frictionless.

Document fraud is on the rise, it is ever more sophisticated, but it can be prevented. Digital markets will continue to expand, as will the opportunities to breach them. Forewarned is forearmed and that is exactly what the next generation of automation can provide and it will only keep improving.


Tags

Related News