What if the “deepfake” contaminated satellite images

The ” deepfake Has become so sophisticated that it is difficult today to distinguish the false from the true. We saw it recently with the videos of a faux Tom Cruise, and the conspiratorial circles feast on false political and scientific interventions to excite the masses. But, what if the phenomenon attacked satellite images with the possibility of manipulating the rendering of aerial maps?

Researchers at the University of Washington thus wanted to show the risks incurred by creating a Artificial intelligence able to modify the appearance of cities. The goal of Bob Zhao, the director of this research published in Cartography and Geography Information Science, is really to demonstrate that it is simple to set up and, ultimately, very dangerous.

Each city has its specificity

For this, the team of researchers trained the AI ​​with satellite images of three different cities: Seattle, Tacoma and Beijing. Seen from the sky, each one has a different appearance depending on the layout of the streets, the greenery, but also the width of the streets. arteries or simply the color of buildings and shadows.

As a result, the AI ​​is able to make the appearance of Beijing in Seattle, and vice versa. When zooming in, it is possible to see inconsistencies, but as for the “deepfake”, the first shoteye is misleading, and who will dig to find out that an image is wrong? According to these researchers, this technique could be used to simulate images of places for which no satellite image is available but also by climatosceptics to demonstrate that the global warming is just an illusion … What’s more interesting, however, is the possibility of bringing old hand-drawn blueprints centuries ago to life.