Are eyes open? Are they blinking?
Many deepfakes have now fixed this one once research exposed this weakness.
Deepfake is an AI-generated video and image manipulation method to create fake videos which are very convincing. Designed to intentionally mislead and spread misinformation. The technology can also create entirely fictional photos from scratch and create voice skins/clones.
So what is deepfake? Really it's the new version of 'Photoshopping' for the 21st century. Using a form of artificial intelligence (AI) called deep learning to make images for fake events so its name evolved as deepfake.
There is so much fake news shared on social media these days but the new trend of deepfake videos can be very realistic and understandably very believable so help spread misinformation. Social media and messaging sites compress videos to make them quicker and easier to share, this removes may critical clues.
Because of the damage, these fake videos can have there has been a challenge set to identify fakes. The Deepfake Detection Challenge (DFDC) was launched by Amazon, Microsoft, and Facebook.
The impact on people with the use of deepfakes, along with other fake news, is the creation of a zero-trust society. No one will know what to trust so will no longer bother to try to distinguish truth from fake news.
Are eyes open? Are they blinking?
Many deepfakes have now fixed this one once research exposed this weakness.
Like all videos look at the quality.
High-definition fake videos easiest to detect as more detail in a video, means more opportunities to see its flaws. But compressed videos for social media help cover these tell tale signs.
Does lip movement match the words that are being projected?
Facial expressions correspond with the exact positioning of the person’s head?
Does skin tone look patchy?
Are there discolorations?
Flickering and bluriness around the edges of transposed face especially where the face meets the neck and hair.
Jewellery, hair, skin and teeth.
Artificially smooth skin
Hair is particularly hard to render well.
Look out for strange lighting effects and inconsistent illumination and reflections.
Slow down the original clip and see if you can see cuts in the video that distorts clips.
Fact-checking the post. Who posted the content? Is the account legit?
Firstly you need to run thousands of face shots of the two people into an AI algorithm called an encoder. The encoder learns and identifies similarities between the faces by reducing them to shared common features, and compressing them.
Then a second AI algorithm called a decoder recovers the faces from the compressed images. Then one decoder recovers the face, and another decoder recover the other.
Swapping faces is as simple as feeding encoded images into the “wrong” decoder.
This is an amazing example of Deep fake in Star Wars of Princess Leia in Rogue One.
Chris Ume Visual effects specialist created a convincing video with VFX and CGI. They were posted on his TikTok account deeptomcruise to show what is possible.
TikTok has actually banned manipulated content back in August to protect its platform against misinformation. They have said that parodies such as Ume’s don’t count so still live.
@deeptomcruise
And bubblegum?!
Sift method is about sorting truth from fiction.
University of Oregon: The SIFT Method: Evaluate Information in a Digital World (Infographic)
Stop: Stop, think and research before liking, commenting or sharing.
Investigate: Investigate into the source
Find: Search for trusted reporting or analysis on the information in front of you
Trace: Trace the quote, quote, or media back to the original source.