Against facial recognition programs that analyze photos, for example posted on social media, tracing the identity of the person in the image, it is possible to make subtle changes in the photos that make them 'indigestible' to artificial intelligences.
This is demonstrated by two apps just presented at the Iclr specialized conference, created by the University of Chicago and Deakin University in Australia and described on the Mit Technology Review website.
The application created by the US university is called Fawkes, a name that recalls the protagonist mask of the film 'V for Vendetta', and acts according to the principle of 'data poisoning', literally 'data poisoning'. The resulting image is indistinguishable to the naked eye from the original, but it confuses recognition programs. In a small experiment with 50 images of all the major programs on the market, from Amazon's Rekognition to Microsoft Azure, the app was 100% successful.
If Fawkes prevents artificial intelligences from recognizing the face in a photo, the application made by Daniel Ma of Deakin University goes further.
The changes introduced in the images manage to completely 'hide' the face, which is not recognized as such by the software.
"This type of technology - explains Ma - can be used to secure your data. It is a new type of defense for your digital rights in the era of artificial intelligence".
(HANDLE).