How to Recognize a Deepfake Without Even Watching a Video
Deepfakes are videos in which one person’s face is glued to another’s body, making the first person look like they are saying or doing something they have never actually done, even if it is something as harmless as Tom Cruise talking with a camera and hitting a golf ball . It’s hard to spot them just by watching the video, but here’s the good news: you don’t have to watch the video to know that you are experiencing a deepfake.
Tom Cruise’s deepfakes (a few short videos posted on TikTok @deeptomcruise ) were compelling because they applied Tom Cruise’s face to an actor who had already built a career as a Tom Cruise wannabe . These videos were professionally filmed and honestly it was impossible to tell that they were deepfakes just by looking at them. Should I worry? Are we living in a world where everything can be faked?
I mean yes, but we’ve been for years. For comparison, consider Photoshop. Of course, the bad one can be obvious. But we usually don’t look at the photos to find out if they’ve been modified. Instead, we just realize that it happens quite often. Do you need to at least look at the cover of a celebrity magazine to realize that the photo is likely altered? You do not.
It’s the same with deepfakes: the context of the video and its origin are important, and not the peculiarities of pixel movement. An expert on videos used for activism told Vice that the bigger problem is that “we live in a world full of petty fakes – simple, incorrectly contextualized or edited videos.”
So how can we detect fake videos without going into every pixel? Mike Caulfield, an expert on digital literacy, recently wrote about it in Twitter , proposing two schemes for the detection of fake anything, and they work well with dipfeykami.
SIFT
The first is SIFT, a concept that he explains here . There are four steps:
- Stop.
- Explore the source.
- Find the best lighting.
- Track claims, quotes, and media back to their original context.
The first step can be applied as soon as you notice that you are addicted, worried or surprised – or when you realize that you are spending too much time examining the image for obvious signs of deception .
The rest of the steps are common sense questions that should be answered clearly if the subject you are looking at is real. Caulfield cited an example from last year where a fake reporter looked convincing on social media. Many details have been checked. But a clear sign of a fake? She said she was writing for Bloomberg, but her articles are not on the Bloomberg website.
Five pillars
Another checklist for verifying video or photographs is called the Five Pillars of Visual Inspection , developed by anti-disinformation organization First Draft.
As they put it:
The good thing about teaching verification is that it can be easily broken. This is because whether you are watching an eyewitness video, a processed photo, a sockpuppet account, or a meme, the basic checks you have to do for them are the same:
Origin: Are you looking at the original account, article, or piece of content?
Source: Who created the account or article, or recorded the original piece of content?
Date: when was it created?
Location: Where was the account created, the website created, or the snippet of content recorded?
Motivation: Why was an account created, a website created, or a piece of content recorded?
As with SIFT, there is no need to enlarge the video or image itself; Instead, you are reducing the scale to assess the context in the real world. This will not only tell you whether it really is, but will also help you understand why it exists and who will benefit from its distribution.
Of course, many of us are already using these methods. For a great example, see Lifehacker’s article in which Nick Douglas investigated the alleged trend of hipsters wearing tiny scarves around their ankles. The image was photoshopped, but in fact, if you follow the links (origin), you will find that it came from an Italian website (source) that publishes humorous articles (motivation). This has never been a real trend.
The SIFT and Pillars methods work just as well for fake and mislabeled news images as they do for silly stories about ankle-length scarves. (Remember when we talked about those “breaking news” social media accounts that spread misinformation – sometimes by accident, sometimes on purpose – by linking it to current events?) Deepfakes and petty fakes exist and you can spot them without any or special actions. learning is just your own common sense if you choose to use it.