A DEEPFAKE (or DEEP FAKE) is a byproduct of machine learning that uses images, audio and video to create a simulated experience. When used deceptively, it becomes a sinister device for easily manipulating and influencing other people. This is done by substituting a person, or their likeness, in an existing image or video with someone else, or vice versa.
The purpose of a deepfake is to spread misinformation and malign a person’s reputation. It is an amalgamation of the words deep learning and fake. Celebrities are often the target of deepfakes, with their images, voices and any other likeness being used to fabricate questionable videos. Actors, recording artists and politicians alike have fallen victim to such scams, underscoring the era of fake news and content generators we live in.
But high profile people are not the only ones who should fear such attacks. Virtually everyone in modern society has some sort of online profile or presence. This makes you an easy target in most cases, especially if your identity has ever been stolen. Compromised identities usually indicate that a person’s financial profile has also been exposed. This means that if you have a significant investment or savings accounts, a hacker just might use ransomware and deepfakes against you for monetary gain.
How to Identify a Deepfake
It can be argued that the first iteration of deepfakes, at least in the Digital age, were images modified with Adobe Photoshop (also known as photoshopped images). We’ve all seen good and poor examples of these. Images altered with graphic authoring tools can still be seen in movie posters, magazine covers and a variety of other contexts. Today this technique is applied to video, which is more difficult to pull of considering the frames per second (FPS) that must be manipulated.
In the early days of deepfake video attacks, one could easily detect their fraudulence through the lack of human behavior. The eyes of the people featured in these videos would not blink because the underlying algorithms weren’t as sophisticated as they are today. But this has since been remedied as specialists have now figured out how to replicate human blinking.
Its becoming more difficult to distinguish between real videos and fake ones, but there are still a few signs to help you do so. Bad lip-syncing and morphed faces that flicker at the edges are perhaps the most immediate. In other cases, the skin tone of the video’s subject is uneven, or detailed renderings, like human hair, appear pixelated or 3D modeled. Teeth and jewelry may be freakishly rendered, and if you look close enough, you might actually see the person’s eyes glinting.
Protecting Yourself against Deepfakes
Everyone has a right to their privacy, but deepfakes have the tendency of destroying this along with a person’s reputation. Anyone could be behind them: An ex-lover, business rival, jealous colleague or just some prankster seeking to harass or extort you. When things take a racy or explicit turn, it is usually considered a case of cyber exploitation.
Developing the skill to identify them is presently the best way of dealing with the problem at large. If you are the subject of a deepfake, being able to prove its fraudulence can go a long way in salvaging your reputation. This, of course, is a reactive way to protect yourself, which is why a concerted effort is being made to automate detection. Researchers at Purdue University, for example, announced in April 2019 the publication of an algorithm that can identify tampered videos. This technology has been made available to everyone free of charge. Check out the video below:
Another way to protect yourself is by keeping your identity in the forefront of your mind. You should use antivirus (AV) products on all your computers and mobile devices to protect your sensitive data. You should also block the phone numbers you’ve received suspicious calls from. For example, if you have ever answered the phone and said…
- Hello? – different inflection
- HELLO! – different inflection
…only to have that call disconnected from the opposite end, chances are it was a ploy to actually record the different tones of your voice so that longer audio files can be simulated using your voice. This is an advanced technique of identity theft.
The growing threat of deepfake content prompted social media giant Facebook to announce its intention of removing media that has been synthesized for reasons other than trying to enhance its quality. Additionally, they will also take down any media produced by artificial intelligence or machine learning that merges, replaces or superimposes content onto video under the guise of authenticity.
Attempts to find an effective solution to deepfake disinformation are being made by experts across the world. In the meantime, your best defense is to be well informed about the threat it poses to your identity and reputation.