Overview

Document based fraud is costing UK business millions (and possibly billions every year). You may think your business is not affected but have you considered the range of possible forms this fraud may take ?

• Insurance claims – fake or exaggerated claims, not only the obvious house or car insurance claims but cancelled flight claims for hotels and a host of others
• Invoices – submitted by suppliers showing inflated costs for goods used.
• Receipts – Are your staff or suppliers claiming a little ‘extra’?
• Identity – many businesses have a requirement to ‘know their staff/customer/tenant/service user’. This is often a legal requirement. What if someone fakes or modifies a document in order to join your organisation, become your tenant or access your services ?

Document fraud is not new, in fact its history goes way back. Remember all those films where some expert forger created fake passports for criminals in a seedy basement ? However, the truth is that technology is making it easier every day to produce high quality false or modified documents, to the point where anyone can use AI and related technologies to do it. Traditionally you took your documents and physically showed them to someone, but now it is all too common to accept scans or photos of these documents being sent via email or uploaded to portals. All of this makes it even easier to make fraudulent submissions.

The demonstration

To illustrate how easy it is to achieve let’s take a look at an example utility bill :

Imagine I am a prospective tenant keen to move into social housing. I happen to know that if the utility bill I am supplying as proof of identity is investigated, the housing association may find information about debts, credit worthiness or other information I would prefer them not to know about. I can use AI or even free tools available on the internet to modify the address on the utility bill and completely mis-direct anyone checking up on me. Once modified, the document gets emailed off to the housing association:

It arrives to a diligent but very busy team who take the document and process it. They perform all the relevant checks and I get approved to become a tenant. Months later, my debts catch up with me and the landlords face a tough time dealing with a tenant that can’t pay the rent !

Now if Klippa DocHorizon was in use, it could check every document for fraud as it was processed. This can be as simple as watching emails as they arrive or processing folders on OneDrive, Google Drive or a range of other options including being call via an API from existing systems. The processing may also extract key information from the documents and feed that into other systems (but more on that in another post). The alert may be as simple as renaming the file to show it has been checked and found to contain potential fraud.

Behind the scenes

Klippa DocHorizon processes the image as it is received and scores the document based o a variety of indicators of possible fraud:

Now the service has identified multiple reasons that this image may be fraudulent:

• Analysis of the meta-data embedded in the image shows that the image was captured at one date/time but actually modified six minutes later. Whilst not proof of tampering it raises suspicion!
• Further analysis identifies that the image was last modified by a tool that could have been used for non-fraudulent purposes such as cropping the image to remove unnecessary borders but again it contributes to the overall risk. (Name of the tool removed)
• The final (and possibly most suspicious) element is that the system has detected ‘splicing’ of the image. That is a high probability that some part of the image has been tampered with. In this case the house number (45) is false and replace the original house number (12).

What Next

At its simplest level the suspicious file may be flagged by renaming it with -ALERT on the end of the filename, it may get moved to another folder or as Klippa DocHorizon becomes more integrated into your other systems it may affect workflows, modify databases etc. In another post I will show how easy it is to use Flow Builder (another component of the DocHorizon product) to easily incorporate fraud detection into your existing systems.

Summary

Some of the more observant readers of this post may have noticed that the first two images (modified and unmodified documents) were deliberately switched round to keep everyone guessing, I also obfuscated the name of the software used to make the edit to hinder those that might want to commit fraud. In reality it would only take your tenants/employees/suppliers a couple of minutes with a popular AI tool to find out how to make that kind of edit, and in reality some AI tools would offer to do it for you. If it is that easy, how do you know some of the documents you are receiving today have not already been modified ?

This is a somewhat simplified example to keep the article short and direct and using sample documents but if you are interested in seeing this demo live, maybe testing it with other types of document and discussing how this low cost option  could fit into your business reach out to me I would be happy to arrange that.

Future

This is only one of a number of articles planned showing how we can plug Klippa DocHorizon into your business systems to use AI to read, intelligently process and enhance documents coming into, going out from and being stored within your organisation.