Interview with Roy Prayikulam from INFORM on the Use of Artificial Intelligence in Combating Financial Crime

At the end of September, Roy Prayikulam was interviewed by wallstreet-online on the topic of Artificial Intelligence and the role it is playing in the fight against financial crime. With permission from the editor, we translated the interview from German to English. The original German version can be read on the wallstreet-online website. The full interview can be read below:

Published by wallstreet-online, Rainer Brosy, September 24, 2020

Fraud is his profession: Roy Prayikulam heads the Risk & Fraud division of the Aachen-based optimization specialist INFORM GmbH. For more than eleven years, he has been fighting financial crime there, together with banks, payment service providers, insurance companies and telecommunications companies using Artificial Intelligence (AI) and algorithms. We asked him about the history of this technology and the daily struggle with fraudsters.

The world of AI has grown rapidly in recent years. How have you managed to establish INFORM internationally as an important player in this field?

For more than 50 years, INFORM has been active in the research and application of software technologies that make and optimize operational decisions that previously could only be attributed to human intelligence. Every year, a part of the profit is reinvested in further research. Our technology integrates methods of Operations Research, Fuzzy Logic and Artificial Intelligence to solve many different planning and decision problems across industries. This can be in logistics, production, supply chain and workforce management or even the fight against financial crime. In this area we have grown strongly in recent years. Two reasons for this are the high degree of flexibility and adaptability as well as the real-time capability of RiskShield, our decision-making system for fraud prevention. After all, in financial fraud, milliseconds count, during which a payment transaction is checked for validity.

Your AI solutions can be applied very well to fraud prevention in the context of the finance and insurance industries. How do you respond to financial crime and what role, in particular, does money laundering play?

Every day, creative criminals develop new methods of fraud. At the same time, in our rapidly changing market, new targets are constantly emerging in the form of new payment methods, FinTechs and services that process sensitive financial data. Nevertheless, our task is to automatically check every payment transaction and every insurance case of our customers for fraudulent activities and to reliably prevent them from being executed, even if only one in a million transactions is fraudulent.

For this purpose, financial transactions and insurance applications are assigned a risk classification based on a variety of different data from different sources, which leads to appropriate consequences: for example, an additional authentication procedure of the account user or a transaction termination. It is important that the algorithmically generated rules and resulting conclusions always remain comprehensible and easily adaptable for the compliance teams. This is also the case with money laundering. When financial institutions work with an AI-based system for risk assessment and fraud prevention, they can take a holistic view of their customer behavior in the search for criminal activity. For example, online and mobile banking activities, cash withdrawals from ATMs, SEPA transfers, customer service contacts and e-commerce transactions can all be examined and related to one another. At the same time, the algorithms also dynamically incorporate external data such as official watch lists from various supervisory authorities.

The Covid 19 pandemic is still a present topic in business and industry. To what extent have you been able to help companies deal with this problem so far?

German banks were urged to provide emergency loans quickly. Many have fallen victim to fraudsters who have misused the relaxed checks to illegally tap into Corona emergency funds earmarked for freelancers and small businesses. Banks around the world are currently struggling with similar waves of false loan applications. But our technology allows us to quickly assess their validity based on IP addresses, device IDs, recipient accounts and other factors so that the money goes to the businesses that are in trouble through no fault of their own, and not the fraudsters. New fraud patterns have also emerged in the areas of e-commerce or phishing, some of which have been reported in the press. But that was already there before the pandemic. Therefore, Covid-19 has helped to draw attention to existing security gaps. A common problem is the lack of cooperation between the departments of financial service providers and the lack of communication between the various tools used in compliance. More holistic and enterprise wide processes should be adopted to best manage risks across all business units and customer accounts.

Software solutions that use AI are versatile and can be used in many application areas. Is there a specific project that has stuck with you over the years?

I remember a particularly audacious fraud attempt a few years ago at a bank within the online banking area. We noticed that some of the transactions we had already successfully identified as criminal were nevertheless debited. The bank's call center had released the blocked payments because the fraudsters had contacted the bank by phone and posed as the legitimate account holders. To do this, they had first spied on the call center's security questionnaire and then obtained the relevant data on their victims - age, birthday, license plate number, name of the first pet, etc. - through social media chats or on dating apps. So, they invested an enormous amount of effort to get the money. Within a few hours, we adapted our fraud prevention model so that transactions released by the call center were stopped in the event of a serious incident and were marked with the instruction to actively call the customer and have the transaction confirmed. This example shows why we cannot fight fraud with rigid, one-dimensional rules, as is still common practice even today in some organizations. We are much more successful if we weigh the burdening and relieving factors in real time to evaluate a transaction. We also need "soft" factors. For this we need AI.

Intensive research on artificial intelligence is currently being conducted in many areas. Which future projects can we look forward to?

At INFORM, we focus on a hybrid approach to AI by combining data- and knowledge-driven methods. The former includes machine learning (ML) algorithms, for example. We search huge amounts of data, in the order of more than 250 million financial transactions daily, for recurring correlations and patterns that indicate criminal behavior. However, if a financial institution has to wait for the training time of the ML models for each new fraud pattern to take effect, the financial damage would be enormous. For this reason, knowledge-driven methods such as fuzzy logic technology are also used, with which human experts define complex rules for dealing with certain behavior patterns. This allows concrete decisions and recommendations for action to be derived even from inaccurate data. In addition, expert knowledge and gut-instinct are incorporated into decision models: an irreplaceable advantage!

---

---

You may also like