Data literacy is our best weapon against fake satellite images

A new algorithm can detect fake satellite images with a high degree of accuracy, but the best guarantee is an informed public.

Cards have always told white lies. To paraphrase the words of Mark Monmonier in his classic book How to Lie with Maps, depicting a curved Earth on a flat plane necessarily involves some distortion.

Fake satellite images are different: the apparent authority of a satellite photograph can make us forget that they have the same vulnerabilities as any other piece of data.

We know there are fake satellite images. The question is whether we can detect them and how reliably. A new machine learning algorithm can spot a particular type of faked satellite imagery with 94% accuracy, but data literacy is the best way to sort reliable images from unreliable ones.

Generative Adversarial Networks (GANs) are frequently used to create compelling deep-fake media. Sometimes the results are less than convincing, as with the recent fake videos of Ukrainian President Volodymyr Zelensky.

Using Cycle-GAN, researchers created fake maps of Tacoma, a city in the US state of Washington. The fake maps included some features of Seattle, Washington and the Chinese city of Beijing.

GANs consist of a generator network and a discriminator network, which work in tandem through cycles of tuning until they have produced a convincing fake based on the characteristics of the data they aim to simulate. .

As the GANs working to produce a convincing mock map of Tacoma went through the tuning process, the map became sharper: shaded areas gave way to simulated roads, detail increased, and intended areas to show land and water have acquired a more natural color.

To the naked eye, Tacoma’s fake map looked genuine.

The machine learning algorithm sorted a deep fake detection dataset consisting of genuine Tacoma satellite images as well as faked images. With a 94% success rate, the algorithm selected the fakes, they were slightly less colorful and had sharper edges than the genuine images.

There is still work to be done to develop the algorithm. It must of course be tested on datasets from other cities. It is effective with CycleGAN images but may not be as effective with other GAN models. And it’s currently limited to a totally correct or totally false binary outcome and it can’t detect when only part of a card has been rigged. Beyond algorithms, however, there is data literacy.

Any satellite image has its limitations as a data source. Images are produced from the top down, so unless a particular satellite has geothermal capabilities, such as synthetic aperture radar (SAR) that can image through clouds, it cannot show that there are people under an opaque structure. For example, a satellite image of the destroyed bridge in the Ukrainian town of Irpin could not indicate that there were people sheltering under it.

Michael Goodchild wrote about citizen sensors in a 2007 article, and this idea is particularly relevant for images of conflict zones. Journalists, volunteers and NGO staff can provide details of what is happening on the ground.

A satellite image is a single data point, but a holistic understanding of a geographic location requires a variety of data sources.

These should come not only from the perspective of a satellite, which in geographic information science is called the eye of the god, but also from the perspective of images of people’s eyes from smartphones, d cameras or drones. Our perception of the conflict in Ukraine can be illuminated by short videos on TikTok and photographs on Twitter. With a range of data sources, we can build a more complete picture.

High-resolution satellite images with superb data quality are available, but they usually come at a cost. US satellite imagery distributor Maxar has provided high-resolution imagery to the Ukrainian government for free, but most high-resolution satellite imagery for public consumption is very expensive. The image updated daily by Maxar from its WorldView-3 satellite was priced at US$22.50 per square kilometer.

To make the most of their resources, most media platforms purchased high-impact images that would appeal to audiences: images of destroyed buildings and bridges, and city blocks reduced to rubble. Few satellite images of the humanitarian corridors have appeared online.

Satellite imagery is still in great demand in order to better distribute humanitarian aid to people forced to flee. More and more NGOs have realized the crucial role of satellite imagery in humanitarian work during conflicts. The free satellite images from have been a vital resource to support the emergency response to the conflict in Ukraine.

It is important that consumers of satellite imagery, both journalists and the public, maintain a critical perspective and examine a range of data sources to assess whether these images are reliable. A fact-checking platform for the public to validate satellite images would also be a big help.

Methods of distorting satellite images will only improve. Ways to keep pace with counterfeiting can be found, but there is no substitute for a vigilant, data-savvy public.

( by Bo Zhao, University of Washington in Seattle)

Previous A deep learning toolkit for viewing and interpreting segmented medical images
Next Google Pixel Watch REVEALED thanks to leaked images