There has been a lot of talk about deepfakes over the past few years. Deepfake is a synthetic media in which a person in an existing image or video is replaced by something else like machine learning and artificial intelligence (i.e. deep learning). In manipulation, visual and audio content can now be created in practice in such a high quality that the viewer or listener cannot distinguish it from the real one. Deepfakes have gained the most attention in making various video content and, for example, fake news, blackmailing, celebrity nudity, stock manipulation and many kinds of frauds.
But it is not only imagine AI-generated people. I just read very a interesting article about geographic deepfakes, which are now not-so-distant threat. In fact, this is an century-old problem, as maps have been falsified for hundreds of years. Now everything is just easy, fast, authentic and automated. Geographic deepfakes can be used in multiple purposes, mostly for misinformation, of course.
In particular, due to the cost of creation, inherently reliable satellite imagery promotes deepfakes. As an example, valuable satellite images with a pixel size of up to 30 cm are available worldwide, from which individual people are already starting to stand out. This makes various deepfake counterfeits very interesting and and also financially attractive.
Deepfake geography may even already be a national security problem, as geopolitical opponents use fake satellite images to mislead enemies or, for example, to hide or betray prohibited sites or objects. A nuclear or weapons of mass destruction plant could be one. Therefore, industry and governments are asked about the methods and technologies or legal constraints for limiting and detecting them. Deepfakes are actually one form of toxic data, which has been more extensively presented in a previous Jarkko´s blog post: #15 What is Toxic data and why should I care?
As in all other areas, raising awareness is important. If we are unaware of the potential for counterfeiting, there is a risk that we will end up with a dystopia of fakes.