The unique pattern of a person's fingerprint can be used not only to identify its carrier, but also to transfer information. It's all about how to interpret the image - researchers at Fudan University in Shanghai have developed a method of encoding information, in which it is disguised as lines on the print. And it is not yet possible to break this cipher.
Any message can be divided into fragments and presented in the form of a polynomial equation, where each term will describe the construction of a curve on a graph. The angle of inclination, curvature, direction of movement, duration - there are so many parameters that the number of their combinations creates a redundant data field to describe any short word, term or set of numbers. After we code our polynomial in the form of curve segments, they need to be transferred to a conditional canvas.
It turns out a chaotic picture of different squiggles and curves, notes the study's author Sheng Li. He noticed that if you introduce certain restrictions, then all these small lines add up to a kind of spiral pattern of a fingerprint. The only thing left is to order the neural network to connect the sections to each other so that characteristic patterns are obtained that will not intersect and contain gaps. For any person, this is just a picture, but if you know which areas to measure and decode, you can exchange encrypted messages.
This technology has two surprising and compelling advantages. Firstly, since the neural network builds all the bends harmoniously, according to well-known algorithms, even if some areas are damaged, you can restore the whole picture and read the message. Secondly, the human eye and fingerprint sensors do not see the catch, because the generated fingerprint is visually identical to the natural one. Whereas data encoding by adding additional pixels to the original picture sometimes unbearably jars the eye and attracts attention with its absurdity.