How to Detect Deepfakes

These days, when you are scrolling on social media and come across an interesting video or image, it can sometimes be hard to know whether what you are seeing is real or AI-generated. Did that celebrity really get married unexpectedly, or did that politician really say that? Or are you being fed AI-generated deepfake content?

This article explores the problem of deepfakes and how to check deepfakes to steer clear of them.

Scam Detectors Most Trusted Websites in Online Security

Guard.io (100): Protect your digital world on any device – Guardio stops scams and phishing in their tracks.
Incogni.com (100): Delete your personal data from the internet and protect against scams and identity theft.
ExpressVPN (100) Stay secure and anonymous online - Best VPN Out There

What Are Deepfakes?

The term “deepfake” usually refers to AI-generated video, image, or audio content that has been designed to mimic a real-life person or scene. They can be created from scratch or by manipulating pre-existing content.

Sometimes they are created as innocent entertainment, as viewers can find out what it would look like if their two favorite animals faced off. But they can also be created with the intention to deceive. These types of deepfakes can be highly damaging both to the person being imitated and the consumer being influenced by fake content.

How to Spot Deepfakes

Deepfakes are swiftly becoming increasingly realistic and widespread, making it difficult for content consumers to determine what is real and what is fake.

While deepfake content is getting better every day, it still tends to have some inconsistencies that you can use to spot them.

Spatial and visual inconsistencies: For example, you can spot some objects that seem out of proportion, and color differences between the actual object as you know it and the manipulated content. Things in the foreground can appear blurry, while objects in the background can be crisp and clear.
Time-based inconsistencies: Video and audio fakes can have time issues, such as a delay between speech and mouth movements.
Face inconsistencies: When deepfakes include people, there are almost always inconsistencies in their presentation. The skin on the face can appear too smooth, there can be incongruous shadows around the eyes and eyebrows, strange glare on glasses or a complete lack of glare or reflection, unfamiliar facial hair not usually seen on the person, not blinking as a typical person would, and the infamous sixth finger on the hands.

While this is a great place to start, recent research suggests that humans are only able to independently detect deepfake content around 55% of the time, which is little better than not at all. Therefore, an examination of the content itself should be matched with consideration such as distribution channels. For example, deepfakes are often circulated on social media by bot and troll accounts, which can often be identified by their metadata and behavior.

Check Deepfakes with Dedicated Software

If you need greater confidence in identifying deepfakes – for example, if you’re a journalist checking the validity of sources – there is software that can detect deepfakes based on their digital signature. Reliable tools include Deepware Scanner, DeepFake-o-Meter, and Optic; however, even these tools are not infallible.

This means that for now, we need to use a combination of common sense and confirmation to identify deepfakes. We can no longer take everything we see at face value.

Verify Another website

Are you just about to make a purchase online? See if the website is legit with our validator:

vldtr^®