The researchers claim that their algorithm is 40% more efficient than previously developed counterparts.
Scientists from Carnegie Mellon University and the Oxford Department of Engineering have developed a special algorithm that helps encrypt secret information in AI-generated drawings, such as DALL-E 2 and the like, writes Interesting Engineering.
The researchers argue that with the growth of tools like ChatGPT and other generative content, you can get more functionality for encrypting files and hiding data, which, for example, are prohibited in totalitarian countries. The algorithm, which everyone can get acquainted with, has already been published on the arXive server, where all the details of working with it are given.
In general, encryption of information in drawings is an old technology known to mankind since ancient times. Of course, with the advent of computers, there is no longer a need for the high skill of an artist who manually draws precise lines, who can later recognize the addressee of the message – special software does all this. And most often this method is implemented to combat piracy, when a special code is embedded in the picture.
Even an ordinary person today can encrypt the text in any photo and send it to the addressee who has the decryption key. There are many programs on the Internet that can do this for many years. And using the free program DeepSound, you can encrypt files even in songs.
Yes, the idea is not new, which is what critics pay attention to, but here it is not so much about the idea of encryption as about a technical solution to this issue. The new algorithm uses a technique called minimum entropy coupling to maximize the overlap between the information distribution of regular content and the content to be encoded so that the difference is imperceptible. The researchers claim that their algorithm is 40% more efficient than the usual ones.
Scientists want their AI encryption algorithm to be available to everyone, because it can be an important tool for journalists and various activists who work in countries with totalitarian regimes.
Earlier, Focus wrote that in China they boasted an algorithm for cracking quantum encryption, but it will not work: why?
