Deepfake The fake that ‘steals’ your face (and privacy)

On 06 July, EACA participated in the webinar “Deepfakes: Are we ready for the upcoming storm?” organised by the European Liberal Forum. The webinar gave a definition of deepfakes, explained the dangers arising from them and tried to give tips on how to recognise them.

What is a deepfake?

Deepfakes are photos, videos and audio created by artificial intelligence (AI) software that, starting from real content (images and audio), can modify or recreate, in a highly realistic way, the characteristics and movements of a face or body and faithfully imitate a given voice. The word ‘deepfake’ is a neologism created by merging the terms ‘fake’ and ‘deep learning’, a particular AI technology. The techniques used by deepfakes are like those of the various apps with which one can amuse oneself by modifying the morphology of a face, ageing it, changing its sex, etc. The starting material is always the “real thing”, i.e.  the real faces, real bodies and real voices of people, but transformed into digital ‘fakes’. Deepfake technologies, developed as an aid to film special effects, were initially expensive and not widespread. But recently, apps and software have started to appear that make it possible to create deep fakes, even very elaborate and sophisticated ones, using an ordinary smartphone. As a result, the prevalence of deep fakes has increased considerably, and with it, the associated risks.

Deepfake and identity theft

Deepfakes are a severe form of identity theft. People who appear in a deepfake without their knowledge suffer a loss of control over their image but are also deprived of control over their ideas and thoughts, which can be misrepresented based on the false speech and behaviour they express in the videos. People in deepfakes may also be depicted in places or contexts or with people they have never met or would never meet, or in situations that could appear compromising. In essence, therefore, a deepfake can reconstruct contexts and situations that never actually happened and, if this is not intended by those concerned, can pose a severe threat to the privacy and dignity of individuals.

How to protect yourself against deepfakes

• Avoid uncontrolled dissemination of personal images or images of loved ones. If you post images on social media, bear in mind that they may remain online forever or, even if you decide to delete them, someone may have already appropriated them.
• Although it is not easy, you can learn to recognise a deepfake. Some elements help: the image may appear pixelated (i.e., a bit ‘grainy’ or blurred); people’s eyes may move unnaturally at times; the mouth may appear distorted or too big while the person is saying certain things; the light and shadows on the face may appear abnormal.
• If you have any doubt that a video or audio is a deepfake made without your knowledge, you should avoid sharing it (so as not to multiply the harm to people by spreading it uncontrollably). And you may perhaps decide to report it as a possible fake to the platform hosting it (e.g., a social media outlet).