<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=565596&amp;fmt=gif">

4 mins read

Digital Fraud

Prevention and detection of document fraud: how Yooz AI technologies change everything

by Jean-Marc Pédréno on 06.17.2020

document fraud detection

In this article, Jean-Marc Pédréno, CTO at Yooz, talks about how Yooz AP Solution helps detecting and preventing document fraud.

 

What are today’s main technological breakthroughs and what are their applications in the finance field?

 

Let us start by saying that protection against fraud is a major preoccupation for our clients. Today, nearly 80% of CFOs say that they are concerned by fraud. When you look at recent studies on the subject, it is understandable to be alarmed: one out of every three companies experienced at least one case of fraud in 2017.

 

We started out by specializing in image processing. We also were pioneers in the field of artificial intelligence, which we have been implementing for nearly 30 years to increasingly automate the processes that we capture electronically.

 

A few words about your R&D ?

 

In SMEs, you often just have “R&D”, or moreover “r&D” with a small “r” and a capital “D”. It is a real advantage for a company of Yooz’s size to have a truly research-oriented team. Ours is comprised of engineers, two PhDs, and 2 doctoral students who concentrate on image analysis and artificial intelligence. Their effort comprises a mini-laboratory that is in charge of research and exploration involving new technologies, as well as developing and industrializing innovative components that are then used by the product development team, with its staff of twenty engineers.

 

Our technical team also includes another twenty people handling operational aspects ranging from support to assistance and solution set-up.

 

Not only that, but our Yooz Lab does not always work alone. The team functions as a true academic laboratory in close collaboration with the scientific community. We have led over 25 collaborative projects with more than 12 European laboratories over the past five years.

 

We are an active member of the scientific community, both for joint publications with our partners – we contributed to over 80 publications in the past 7 years; and by helping to organize major international conferences in the field of document analysis, such as ICDAR (International Conference on Document Analysis and Recognition) and DAS (Document Analysis Systems). Our researchers have over 25 years of experience working on academic collaborations and are the authors of over 50 publications. 

 

We recently reinforced our ties with L3i, the University of La Rochelle computer science laboratory, by setting up a Yooz branch in La Rochelle. We are also working on other major collaborations, such as the Feder SECURDOC project on document fraud, supported by the Nouvelle Aquitaine regional authorities. This year, we created a joint Yooz-L3i laboratory named IDEAS with support from the French National Research Agency. This laboratory enables us to share resources on research topics that are common between Yooz and L3i.  

 

That is the context in which we decided to expand our thinking about functionality to apply for invoices and the Yooz solution overall, combining the power of artificial intelligence with our expertise in image processing.

 

Our goal was also to be able to dive as deeply as possible into false document detection and take action on all types of fraudulent operations.

 

Our solution goes beyond fraud detection technologies and also integrates a reliable audit trail and archives with probative value, enabling us to offer an all-in-one module for security and supplier process protection for all companies.

 

More precisely, what technologies were developed to fight against fraud?

 

We use a variety of different artificial intelligence technologies that enable our clients to make their processes more secure and more reliable.

 

These expert systems are involved from document capture, the very first step in the document automation process, to detect fraudulent types of behavior. Examples include using steganographic shape and metadata analysis to detect modified information on an invoice by tracking down any changes made in the image, such as bank details, which is one of the leading sources of invoice fraud. Steganographic methods include a frequency analysis component that detects falsified information by identifying changes made to an image, such as amounts, dates, or names.

 

These technologies integrate powerful algorithms that can not only adapt to all types of documents – even those with variable structure – but also learn from examples and create their own knowledge base. For example, statistical analysis can detect monetary amounts that are out of the ordinary with respect to known data in supplier history. Solution users can then be alerted if amounts higher than usual are found. This type of technology also detects suspected duplicates, that is, invoices with the same numbers and supplier names.

 

With Yooz, companies are able to identify fraudsters easily. The module provides complete traceability of all interventions, indicating dates, modified values, and identifiers of the person who made the changes.

 

 

What is a concrete example of fraud detection or a fraudulent operation where your intervention is possible? 

  

Our technologies are used to detect falsification on various types of documents, such as pay slips, identity documents, proof of address, etc.

 

Pay slips are a very real example. By implementing steganographic analysis technologies, we can detect whether information such as numbers, dates, names, or salary amount were modified. 

 

Find out how you can protect your business from document fraud!

 

Download our whitepaper