To do this, the neural network collects data from the accelerometer, light sensor and other sensors of the mobile device.
A team of scientists from Nanyang Technological University in Singapore has developed a neural network that can guess the smartphone’s pin-code with an accuracy of 84%, writes N+1. The only information the program needs to find the right password is the data from the accelerometer, light sensor and other sensors that are present in most devices.
Scientists have written a special mobile application that is able to collect such information — the user’s permission is not required for this. The basic idea is that the position of all keys on the keyboard is known, which means that by tilting the smartphone, changing light and other data, you can calculate which key was pressed.
A neural network, on the basis of which the program operates, is needed in order to process the entire volume of data collected from sensors and calculate the necessary combination of numbers. When testing the program it was found out that on all possible combinations of four digits, the recognition accuracy at 20 attempts was 83.7%. The accuracy at recognition among the 50 most common pin-codes is 99.5% from first attempt.