Boffins at Fujitsu and Japan’s Nagoya university are claiming to have successfully developed technology designed to prevent phone scammers by recognising certain keywords and detecting changes in voice pitch and level.
The technology was developed as part of the "Modelling and Detecting Overtrust from Behaviour Signals" research area led by Kazuya Takeda.
It focuses on the notion of “overtrust”, the situation that occurs when a human is overwhelmed with distressing information and loses the capacity to objectively evaluate whether they are being lied to or not.
The technology analyses the pitch and volume of the potential victim’s voice to detect when a situation of overtrust is occurring. Typically a person's voice flatens out in the high frequency range when they are put under psychological stress, and from this Fujitsu said it can can infer a situation of overtrust with over 90 per cent accuracy.
Bolstering the detection capabilities is keyword detection functionality which uses a list provided by the National Police Academy and counts the number of times any of those keywords are spoken, ignoring any other words.
The tech will then make a decision on whether the potential victim is being scanned by assessing the number of keywords in the conversation and whether their voice indicates a situation of overtrust, said Fujitsu.
This may seem like a lot of bother to be going to for a crime which rarely hits the front pages in the UK, however phone scams are a big problem in Japan and deemed particularly reprehensible in a country where crime is rare as it is usually targeted at the elderly and infirm.
Typically the scammer will pretend to be either a member of the victim’s family, friend or someone in a position of authority. They will usually impart some distressing information about a crime committed by an acquaintance of the victim and urge them to send money to sort the problem out.
Fujitsu now plans to work with the National Police Academy and The Bank of Nagoya to test the technology embedded in mobile phones. ®