With the click of a mouse, AI knows when you’re lying online

12 Jun 2017

Image: Sergey Mironov/Shutterstock

Think you can fool people online? With the simple click of a mouse, new AI software might be able to see right through any deceit.

Online identity theft is a serious issue that isn’t going away anytime soon, with research showing that in the US alone, 15.4m people had $16bn stolen from them in 2016.

To make matters more difficult, those conducting this fraud are using more and more sophisticated methods to trick services into thinking they are dealing with the actual account holders.

So, in an attempt to stop fraud at its source, a group of Italian researchers went about developing an artificial intelligence (AI) algorithm that has shown an ability to identify when fraud is taking place, with just a few clicks of a mouse.

According to Science, the AI software would pose a series of personal questions to the user that are answered either true or false, and their reaction time is then measured through the movement of the mouse.

The 12 questions are designed to be easy for the actual user, such as the year they were born in, or even ‘Is Giulia your name?’

The researchers then measured a number of responses from users and saw how long it would take people to click ‘yes’ or ‘no’ in the top two corners of the screen.

To add another layer of complexity, 12 further unexpected questions were asked of the participants, including their zodiac sign, in an attempt to prevent potential fraudsters from rehearsing basic information.

The software was then asked to figure out which of the study’s participants were responding truthfully and which were posing as someone else, resulting in some interesting findings.

A proof of concept

Four different algorithms took part in the experiment based on their answer times alone, resulting in a high score of 85pc accuracy.

However, when mouse paths and clicks were introduced into the equation, this shot up to a high score of 95pc.

Discussing its future possibilities, researcher Giuseppe Sartori said the new AI could be used as a first-screening process for various online services, as a good proof of concept to take forward for further development.

One hindrance, however, is its feasibility in a future where mouse clicks are replaced by finger taps as millions of people switch from desktop to mobile devices.

Last year, a study claimed that for the first time, the number of mobile users exceeded desktop users, with 51.3pc of webpages loaded on mobile.

Colm Gorey was a senior journalist with Silicon Republic