How to detect a deepfake without even watching the video

Real Tom Cruise, for comparison

Real Tom Cruise, for comparison
Photo: Emmanuel Wong / Getty Images for Paramount Pictures (fake images)

Deepfakes are videos that stick the face of one person to the body of another, making the first they seem to be saying or doing something they never did, even if it is something as harmless as Tom Cruise talking to the camera and hitting a golf ball. They’re hard to spot just from watching the video, but here’s the good news: you don’t actually have to watch the video to know you’ve encountered a deepfake..

Tom Cruise’s deepfakes (some short videos posted on TikTok on an account called @deeptomcruise) were convincing because they applied Tom Cruise’s face to an actor who had already built a career as a Tom Cruise impersonator. These videos were professionally produced and honestly there was no way to tell them they were deepfakes just by looking at them. Should we be worried? Do we now live in a world where anything can be faked?

I mean yeah, but we’ve been for years. For comparison, think Photoshop-retouched images. Sure, a bad one may be obvious. But we do not normally examine photos to find out if they have been altered. Instead, we are aware that it happens quite frequently. Do you need to match? Look on the cover of a celebrity magazine to find out that the photo is probably altered? You do not.

The same goes for deepfakes: Wwhat matters is the context of the video and where it came from, not the details of how the pixels move. Expert in videos used for activism told Vice that the biggest problem is that “we live in a world with many superficial fakes: simple videos, badly contextualized or edited”.

So how can we detect fake videos without carefully studying each pixel? Mike Caulfield, digital literacy expert, tweeted about this recently, offering two frames to detect counterfeits anythingand they work fine for deepfakes.


The first is SIFT, a concept he explains here. There are four steps:

  1. Stop.
  2. Investigate the source.
  3. Find better coverage.
  4. Trace claims, quotes, and media back to the original context.

The first step can be applied as soon as you notice that you are being absorbed, worried or wondering, or when you realize that you are spending too much time scrutinize an image to telltale signs of deception.

The rest of the steps are common.Feel the questions that should have clear answers if what you are looking at is real. Caulfield pointed to an example from last year of a fake reporter with a compelling social media presence. Many details were reviewed. However, the obvious sign of counterfeiting? He said he wrote for Bloomberg, but his articles do not appear on the Bloomberg website.

The five pillars

Another checklist for watching a video or photo is called Five Pillars of Visual Verification, from the anti-disinformation organization First Draft.

As they put it:

The nice thing about teaching verification is that it is easy to break. This is because whether you are watching an eyewitness video, a tampered photo, a puppet account, or a meme, the basic checks you need to run are the same:

Origin: Looking at the account, article, or original content?

Source: Who created the account or article, or captured the original content?

Date: When it was created?

Location: Where was the account established, was the website created, or was the content captured?

Motivation: Why was the account established, the website created, or the content captured?

As with SIFT, there is no need to zoom in on the video or the image itself; instead of, you get closer out of to judge your context In the real world. That Not only will it give you clues as to whether it is real, but it will help you think about why it exists and who will benefit from it being shared.

METERany of us already use these techniques, of course. For a perfect example, see this piece from Lifehacker, in which Nick Douglas investigated the alleged tendency of hipsters to wear small scarves around the ankles. The image was a photoshop, but the truth is that if you follow the links (provenance), you find that it comes from an Italian website (source) that publishes humorous articles (motivation). It was never a real trend.

The SIFT and Pillars techniques work just as well for mislabeled, fake news images as they do for silly ankle scarf stories. (Remember when we talk about those “breaking news” accounts on social media that circulate misinformationsometimes by accident, sometimes on purposelinking it to current events?) Deepfakes and shallowfakes are out there, and you can spot them without any special training, just your own. common sense, if you decide to use it.


Source link