DeepFake: the 4 tips for detecting fake videos

detect deepfake
detect deepfake

Have you ever heard of deepfake? This fascinating, yet disturbing technology makes it possible to create manipulated videos so convincing that they can fool even the most discerning eyes.

In our previous articles, we explored how these fake videos are produced. Today we are going to focus on an equally important topic: how to detect deepfakes.

Whether you're a techie or just concerned about the authenticity of the videos you watch online, you won't want to miss the techniques we have to offer.

1) Eye blink

The first defect of deepfakes that can betray the falsity of the video: blink abnormalities. This detail can be difficult to detect at first glance, but software developed by researcher Siwei Lyu makes it possible to specifically analyze this characteristic and detect any conclusive anomaly.

Fake videos often have one thing in common: characters who blink much less frequently than real people. Indeed, humans blink their eyes every 2 to 10 seconds when they speak.

If you notice a flaw in the blinking rhythm in a video, it could be a sign that the video is fake. Don't be fooled by deepfakes! Keep an eye out for eye blink anomalies to detect fake videos.

Authentic video

Rigged video

2) Spotting facial deformity

In addition to eye blinking, deepfakes can also feature facial distortions that may reveal the falsity of the video. Strategic areas of the face, such as the nose, eyes, and mouth, are most likely to contain abnormalities that betray video manipulation.

MesoNet software has been specially designed to detect these types of irregularities in strategic areas of the face.

Using careful analysis, MesoNet can spot distortions, inconsistencies in facial colors, and other anomalies that may indicate the presence of a deepfake.

3) Audio analysis

In addition to visual analysis, audio analysis can also help spot deepfakes. Rigged videos may have audio anomalies, such as voices that don't match lip movements or inconsistent background noise. Here are some tips for spotting fake videos through audio analysis:

  • Verification of lip movements : If the lip movements do not match the lyrics in the video, this may indicate audio manipulation. Listen carefully to spot inconsistencies.
  • Background noise analysis : Deepfakes may also contain inconsistent background noise. If you hear sounds that don't match the context of the video, it may be a sign of a faked video.

4) Source Verification

Finally, the last technique is the source verification. If the video comes from an unknown or dubious source, it may be fake. Also search for the same video on other sites or sources to see if it has been flagged as fake or suspicious.

How can we protect ourselves from deepfake?

New emerging software offers proactive protection against deepfakes by authenticating videos as they are taken. One of these tools is Amber Authenticate. It is a tool that uses the blockchain Ethereum to generate scrambled cryptographic hashes while shooting.

This allows the authenticity of the video to be confirmed at any time after shooting. If the video is authentic, Amber Authenticate adds a green frame around the video. However, if any part of the video is modified, the application detects this modification and the frame changes from green to red.

By using tools like Amber Authenticate, you can take proactive steps to fight deepfakes and protect your information and reputation from fake videos online.

The Amber Authenticate app
The Amber Authenticate app

In sum, to combat deepfakes, it is important to take preventative measures such as using apps like Amber Authenticate to help authenticate videos as soon as they are taken. With the help of these new tools, we can work together to reduce the spread of fake videos on the Internet.