Emergence of Deepfakes
Deepfakes are realistic videos that apply artificial intelligence (AI) to depict someone say and do things that never happened. These technologies make it increasingly difficult to distinguish between real and fake media. Harnessing the generative brilliance of deep learning techniques, deepfakes merge, replace, and superimpose images and video clips to create fake videos that appear authentic.
While early examples of deepfakes focused on political leaders, actresses, comedians, and entertainers having their faces weaved into porn videos or them saying and doing things they didn’t, deepfakes in the future will likely be used for revenge porn, bullying, fake video evidence in courts, political sabotage, terrorist propaganda, blackmail, market manipulation, and fake news.
A combination of “deep learning” and “fake”, deepfakes rely on neural networks that analyze large sets of data samples to learn to mimic a person’s facial expressions, mannerisms, voice, and inflections. The process involves feeding footage of two people into a deep learning algorithm to train it to swap faces in videos.
Although deepfakes usually require a large number of images to create a realistic forgery, researchers have developed a technique to generate a fake video by feeding it only one photo such as a selfie. This enables movie makers to recreate classic scenes in movies, create new movies starring long-dead actors, and make use of special effects and advanced face editing in post-production.
Brands can contract supermodels who are not really supermodels, and show fashion outfits on a variety of models with different skin tones, heights, and weights.
Creators of Deepfakes
There are many types of deepfake producers: communities of deepfake hobbyists, political players such as foreign governments, and various activists, other malevolent actors such as fraudsters, and legitimate actors, such as television companies.
Many hobbyists focus on porn-related deepfakes, while others place famous actors in films in which they never appeared to produce comic effects. Mostly, hobbyists tend to see AI-crafted videos as a new form of online humor, and contribution to the development of technology to solve intellectual problems, rather than as a way to trick or threaten people.
In the light of modern cyberwarfare, deepfakes weaponize disinformation to interfere with elections, skew voter opinion, and breed civil unrest. Domestic and foreign state-funded internet “troll farms” use AI to create and deliver political fake videos tailored to social media users’ specific biases.
Deepfakes damage
The most damaging aspect of deepfakes may not be disinformation, but rather how constant contact with misinformation leads people to feel that much information, including video, simply cannot be trusted, thereby regarding everything they see as deceptive.
Criminals have used AI-generated fake audios to impersonate an executive on the phone asking for an urgent cash transfer. In the political scene, a 2018 deepfake created by Hollywood filmmaker, Jordan Peele, featured former US President Obama discussing the dangers of fake news and mocking the current president Trump. In the future, video calls will be faked in real-time.
Combatting Deepfakes
There are four ways to counter deepfakes: 1) legislation and regulation, 2) corporate policies and voluntary action, 3) education and training, and 4) anti-deepfake technology that includes deepfake detection, content authentication, and deepfake prevention.
Media forensic experts have suggested subtle indicators to detect deepfakes, including a range of imperfections such as face wobble, shimmer and distortion, waviness in a person’s movements, inconsistencies with speech and mouth movements, abnormal movements of fixed objects such as a microphone stand, inconsistencies in lighting, reflections and shadows, blurring of facial features, unnatural eye direction, missing facial features, misalignment in face symmetry, inconsistencies in pixel levels, and more.
AI is instrumental in detecting deepfakes. AI can either look at videos on a frame-by-frame basis to track signs of forgery, or review the entire video at once to examine soft biometric signatures, including inconsistencies in the authenticated relationships between head movements, speech patterns, and facial expressions such as smiling, to determine if the video has been manipulated.