Microsoft’s New Deepfake Tool Can Detect AI-Manipulated Media


INTRO: With more and more use of AI (AI) on pictures and videos, “deepfake” content has become quite common on the web . Although many of those content is formed for fun and entertainment, these can still become sources of misinformation for the less-informed population. So, to battle the spread of misinformation and faux news via “deepfake” images and videos, Microsoft has made a tool which will detect images and videos that are AI-made.

The “Video Authenticator” tool, developed by the R&D division of the Redmond-based software giant, can analyze images and videos and detect if those are artificial or real. The tool studies the videos and therefore the images and provides “a percentage chance, or confidence score” to them.

“Today, we’re announcing Microsoft Video Authenticator. Video Authenticator can analyze a still photo or video to supply a percentage chance, or confidence score, that the media is artificially manipulated.”, wrote Microsoft’s executives in a politician blog post.


According to Microsoft, just in case of analyzing videos, the tool will provide the “percentage chance” for each frame of the video in real-time.

Although detecting AI-manipulated media can tough, the corporate says that the tool “works by detecting the blending boundary of the deepfake and subtle fading or greyscale elements which may not be detectable by the human eye.”

Now, coming to some technical details, the “Video Authenticator” tool was developed by employing a public dataset from FaceForesnics++. it had been then tested on the DeepFake Detection Challenge Dataset. And as per Microsoft, these two are the “leading models for training and testing deepfake detection technologies”.

The Redmond-based giant is collaborating with an AI Foundation of San Francisco to distribute the tool. the corporate , with the introduction of this tool, aims to stop the spread of misinformation, especially before the US elections.

So, the tool “will initially be available only through RD2020 [Reality Defender 2020], which can guide organizations through the restrictions and ethical considerations inherent in any deepfake detection technology.”. Hence, it’ll be distributed to election campaign organizers and news and media outlets covering political news.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.