The computer learned to recognize the gender of a person by using the keyboard

The type of typing on the keyboard is different for everyone, but all men and all women have common features that allow you to set the floor with confidence greater than 95 percent.

For their work, Ioannis Tsimperidis (Ioannis Tsimperidis) and his colleagues from the Greek University of Democritus selected 75 volunteers (36 men and 39 women). On their computers were installed specially created programs-keyloggers IRecU, which within 10 months fixed each pressing of the key.

The collected files were analyzed by the statistical analysis program ISqueezeU. She allowed to assess the possible correlations of a person's sex with different features of typing: the time of keystrokes, pauses between individual letters, and so on.

The analysis showed that some of these characteristics are indeed significantly correlated with the sex. In an article published by Digital Investigation, scientists among such features are called the pause between pressing the N key and the next O, as well as M and O. The reasons for these differences were not of interest to the authors.

These same characteristics and associated weights were used to train five different neural networks, which were soon able to determine the gender of the user with an accuracy of more than 78 percent, with the success of the most successful of them exceeding 95 percent.

The authors believe that such an analysis can become a tool for an inconspicuous and unobtrusive definition of a user's sex - for police investigations or planning targeted advertising campaigns.

naked-science.ru