Feedzai: Portuguese scientists have developed an advanced method for detecting banking fraud, based on experience they had gained in creating space flight simulation software.
Feedzai's software is used to monitor every electronic market in Portugal, while at a global level the company's tools oversee payments of $229 billion annually.
The space missions and fraud detection systems have one thing in common: the need for highly advanced software. «When a spacecraft is launched it needs software for its guidance» explains Paolo Marques, who was a consultant for the European Space Agency (ESA) before founding Feedzai in 2009. «You also need software for communication with the ground».
Long before a vessel is launched, the software must be thoroughly examined for issues. However, there is a problem, as Paolo points out: «you don’t have the spacecraft yet».
So, the scientists are building a «universe» of software to simulate the mission. «What you need is something that represents the spacecraft, the mission control and the ground stations, as well as many other elements, so you can control them all».
At ESA, Paolo and Nuno Sebastiano of Feedzai were using high‑performance computing techniques to create virtual satellites: «computer clusters that pretend to be whatever is involved. A computer acts as a spacecraft».
The software must be very stable so that it can perfectly emulate every element of the mission and the vessel, and it must be able to do it quickly. This software is also used for operator training.
Similar challenges are also encountered in fraud detection: for example, in both cases processing large volumes of data in real time is required, and the software must be able to recognize anything that moves outside the sphere of the normal.
However, a significant difference is that in the case of space missions there are some fixed and exception‑free rules, whereas in fraud detection decisions must be made on a case‑by‑case basis: a temperature change in a spacecraft is always a problem, however each bank customer entails different habits and data.
Therefore, the software must be able to detect what is normal for an entrepreneur and what for an educator before it flags «strange» behavior.
Feedzai developed an artificial intelligence system that can process a large volume of transactions, checking each transaction made by a client over a period of four years. Using machine learning and big data techniques, the software «learns» to distinguish the «suspicious» from the «non‑suspicious» transactions.
Feedzai's software monitors over 300 variables per individual, creating very detailed and personalized spending profiles. «Overall we monitor over five billion variables simultaneously» says Paolo.
«It's like having 500 extremely smart people monitoring every individual transaction and deciding based on their experience whether it's fraud or not. It's a massive amount of computing power» adds.
According to Carlos Serqueyra of Instituto Pedro Nunes, the technology to save money for banks, while also improving the so‑called customer loyalty: «Feedzai’s machine learning models and big data science are able to detect fraud 30% earlier compared to traditional methods and demonstrate how capabilities developed in ESA research centers can be useful in other areas».
According to the relevant announcement by ESA, each year 11,4 billion dollars are lost due to credit card fraud- number expected to rise as cybercriminals evolve their methods.
