Making cyber fraud detection methods quicker and more reliable

Optimal stopping theory: applying probability and statistics to fraud detection

"In current methods, the normal card user pattern of expenditures is constructed and then subsequent expenditures are analyzed to determine if there are deviations," explains Dr. Bruno Buonaguidi. "In our method, based on optimal stopping theory, the expenditures pattern of a card user is monitored constantly over time. When the probability of having a fraud exceeds a certain threshold, then we can declare its occurrence." To illustrate how optimal theory works, Dr. Bruno Buonaguidi uses the following comparison: "Imagine you are observing the evolution of a stock price. You want to know when to sell or buy the stock in order to make a profit. What you need to be looking at is ‘the change point’, the time at which a change occurs in the features of the stock. That way, you are able to react promptly and save or make money. By using optimal stopping theory, you can infer that ‘change point’ by monitoring the stock market data as it is collected. Coming back to our context, the same goes with the expenditures of card users."

Working on a theoretical model is the first step in Dr. Buonaguidi’s research plan. The second will be to develop a technology that will be used together with current methods for the detection of cyber fraud. The ultimate goal is to contribute to the challenging problem of rapidly and correctly disclosing frauds in electronic transactions. In this sense, Dr. Bruno Buonaguidi’s research could have an important impact on society. Indeed, the negative consequences of frauds in electronic payments are not only suffered by banks, but also by merchants, card users and insurance companies. Despite the progress made in the field of electronic fraud detection, especially artificial intelligence methods, the current cost of electronic fraud remains extremely high and is expected to increase in the future. Current methods would benefit from an additional instrument capable of disclosing fraud more rapidly, but also more efficiently to avoid unjustified blocks