Subscribe
About
  • Home
  • /
  • Security
  • /
  • AI-powered document-forgery detection fights shallowfakes

AI-powered document-forgery detection fights shallowfakes

Staff Writer
By Staff Writer, ITWeb
Johannesburg, 01 Sep 2022

Skyrocketing fraud continues to plague the insurance, banking, retail, telco, and gaming sectors, with the spotlight shining on shallowfakes as a growing threat alongside more sophisticated deepfakes.

While deepfakes require AI to create realistic but synthetic still or video images or voice recordings, a shallowfake is a method of manipulating media content without the use of sophisticated AI or machine learning technology and algorithmic systems, using instead simple video editing software.

Clive Gungudoo, director of financial crimes and risk management at MoData, says the accessibility of tools to modify documents, from IDs, bills, bank statements, photos and more, means that it is becoming easy for criminals to alter documents and images at speed and on scale.

A rise in shallowfakes

Document fraud is nothing new, but it is the scale of the problem that is unprecedented, Gungudoo says.

A slew of factors have contributed to this rise, particularly the increase in self-service automation, he adds. “With the digitalisation of the insurance and banking industries, hard copies of documents are rarely required as organisations now rely on existing and new customers to either send digital documents and images via email or to upload them themselves using the organisation’s online portals.”

In addition, the COVID-19 pandemic accelerated self-service across many industries through necessity, and customer on-boarding is now largely done online rather than in person.

This dramatically increases the opportunities for bad actors to alter and manipulate documents digitally prior to upload. They tend to use ‘synthetic identities’ – a mix of fake, stolen or data-breached personally identifiable information (PII) that check out with third party data sources.

“This modus operandi sees criminals using ‘mules’ – real people with real identities – to bypass secure biometric and identity verification during remote onboarding and account opening,” says Gungudoo. “The uploaded digital documents are generally not checked during the automated online processes, allowing fraudsters and money launderers to get through the compliance gates.”

Cost of living

He adds that the cost-of-living crisis also plays a key role in the rise of opportunistic fraud.

When people are under financial pressure, they are more likely to tamper with invoices and supporting documents to, for example, inflate an insurance claim or to secure loans and financing. They might operate with fraudulent intent and falsify electronic bank proof of payment advices in exchange for goods and services, a tactic also used by large-scale fraudsters, he explains.

“Similarly, opportunistic occupational fraud is increasing. Often in these cases, procurement invoices are forged, amounts are inflated, and payments are made to personal accounts of staff or family members.”

He says that in large-scale fraud operations, document forgery is prolific, particularly in relation to invoice redirect scams and business email compromise fraud. “In these cases, fraudsters intercept invoices and change the banking details before they are sent for payment processing. The electronic payment then ends up in mule accounts controlled by the fraudsters.”

Using AI to detect shallowfakes

In the past, Gungadoo says document authenticity checks were conducted manually, but the subtlety of shallowfake makes them defy all but the closest scrutiny. 

The solution, he says, is AI-powered detection. While AI is not needed to create shallowfakes, it is extremely useful in detecting them and helps to counter next-generation fraud and money laundering, where growing mule accounts are at the centre of facilitating financial crimes.

MoData’s AI-powered document authenticity solution means documents are automatically scanned on submission, picking up alterations to images and documents in virtually real-time.

Acceptance or the red flagging of documents, including PDFs, JPEGs, PNGs, and TIFFs, is automated, saving both time and resources.

“Document forensics is also able to detect forgery patterns across documents, thwarting tech-savvy fraudsters who might be manipulating the same document several times or using sophisticated templates,” ends Gungadoo.

Share