Alex Graves, a specialist in Long Short-Term Memory recurrent neural networks at the University of Toronto, published a number of fascinating findings before the documentation was quickly removed. Graves’ research is astonishing and revolves around the concept that specific data points in combination with a Long Short-Term Memory recurrent neural network can accurately predict complex sequences. To test this hypothesis, Graves created a program which analyzes the graphology of handwriting.
The software is able to judge the characteristics of each person’s handwriting technique and formulate new sentences with a staggering degree of accuracy. It’s not perfect, and a professional graphologist could probably tell the difference. However, it’s a technical feat and one which could be used to forge your signature or personal style. Everyone’s handwriting contains an assortment of individual nuances and computers can now decipher between minute changes in cursive text. The research team have released the tool which converts up to 100 characters in 5 distinct handwriting styles. In all honesty, I can’t see this being used in a positive manner and only to help online fraud. This probably explains why the original research post was removed from all public avenues.
Do you think this technological feat will simply be used be cybercriminals to commit identity fraud?
Thank You Tom’s Hardware for providing us with this information.
