DeepFakeDetection

• 

  Logo

• 

  Workflow

Inspiration

Creating deepfake videos is getting easier and easier. You don't need technical skills anymore to make a manipulated video. You just follow an online instruction, that's enough.
At the same time, the videos are becoming more and more realistic.
Humans are far inferior to AI when it comes to recognizing deepfake videos. Therefore, it is now possible to mainipulate entire societies, since the naked eye can no longer recognize deepfakes.
The question for everyone is how to protect themselves and others from this disinformation.

Part of the solution is to critically question and check if what you see is plausible. In addition, however, we need technical tools and have to fight AI with AI.

What it does

This is a Chrome browser extension to detect deepfakes in YouTube videos.
The user has the possibility to record a suspicious part of the video. The recording is automatically sent to the AI for analysis.
The result of the AI analysis is displayed to the user.
The user can also download the analyzed video.

Example:

Example output for no deepfake:

How we built it

This is a Chrome extension that use the new Manifest V3 version.
The programming language used is JavaScript, HTML, CSS.

Challenges we ran into

Recording the video sequences was not easy. In addition, the MediaRecorder does not support mp4 format. Here Brad from Modzy helped me a lot and extended the model to handle webm formats. Also, it wasn't easy to always show the extra button in the player because YouTube loads dynamically.
There were also processing issues that Brad was able to fix.
In general, I was limited by the chosen architecture (chrome extension) in the possibilities to use the existing SDKs. Very helpful was the CLI support.

Accomplishments that we're proud of

In the Chrome Webstore I searched in advance for deepfake detectors and found nothing. So this extension will be the first possibility to check deepfakes in an uncomplicated way.

What we learned

Modzy was completely unknown to me before this project.
I learned a lot about the SDKs and the APIs provided.
I also learned a lot about the provided models.
How urgent the deepfake issue has become for the society.

What's next for DeepFakeDetection

When version 0.0.7 is available I will finally implement the popup window. There it will show a history of the previous analyses with link, description, AI evaluation.
Currently only a mockup is displayed.

I could not test this until the hackatron was finished, because version 0.0.6 occasionally runs into an error.
After that the extension will be published in the Chrome Web Store.

Because the software will evolve, I have created a release here:
Release v1.0

The zip can be downloaded and the /dist folder can be loaded as a Chrome extension.
Here's how:

• In Chrome open: chrome://extensions/
• Enable developer mode
• Load the /dist folder
• Enter API key
• Open YouTube
• Watch a video if you don't see the new icon in the player, please restart chrome once
  

For the current software status, see the GitHub link below.

Built With

• chrome
• css
• html
• javascript
• modzy