Researchers at Duke University created a virtual colony of bacteria and then developed a coding scheme based on it, which they called emorfi, reports sciencedaily.com.

Scientists have settled virtual bacteria in a virtual Petri dish to be able to control their number and growth. Depending on certain conditions, like nutrient levels and limited space, bacteria grow in a certain way. If you control this growth, you can create something like a code.

“The encoding cannot be called ‘specific’ because the simulated patterns corresponding to certain letters are not always the same. However, the researchers found that a machine learning program can learn to distinguish between them in order to recognize a given letter,” the material says.

“You can look at the photo of me for some time, but none of the images will be an exact copy of the other,” explained the principle of AI work on the “bacterial” code Lynchong Yu, a professor of biomedical engineering at Duke University. “But if all the images constantly emphasize how I usually look, you can easily recognize me even if you are shown a photograph of me that you have never seen before.”

To encrypt messages, the encoder must create a series of patterns, each corresponding to a different letter. Although the patterns may appear similar to each other to the naked eye, a computer algorithm can distinguish them.

Scientists argue that a code made from bacteria is extremely difficult to crack if the cracker does not know the initial conditions for the growth of bacteria, and powerful AI will be needed to try to determine them.