In an effort to combat deepfakes, Microsoft has released a new tool verification video, which can analyze a photo or video and show if there's a chance it's been tampered with (is fake).
In the case of a video, Microsoft said the tool will provide a real - time percentage for each frame of the video (as it plays) that indicates the likelihood of it being a deepfake. The tool detects changes in colors, fading, and grayscale elements that may not be detectable by the human eye.
Deepfakes can be photos, videos or audio files that are processed by artificial intelligence (AI) tools. Microsoft said that detecting deepfake files is crucial, especially now that we are so close to the election US.
The new technology was created using a public dataset from Face Forensic++ , and Microsoft said it was tested on the DeepFake Detection Challenge Dataset , which is the leading model for training and testing deepfake detection technologies
“We expect that methods for creating deepfakes will become more sophisticated.” However, Microsoft said that all AI detection methods have failure rates, so we need to understand that and be prepared to respond to deepfakes that bypass detection processes.
“So, in the long term, we need to look for stronger methods for preserving and certifying the authenticity of news articles and other media“.
Some tools have already been created for this purpose, but Microsoft also unveiled a new technology that it said can detect tampered content and reassure people that the media and files they are watching are authentic.
This technology has two components: the first is a tool built into Microsoft Azure, which allows a content producer to add digital hashes and certificates to a piece of content.
“Hashes and certificates then ‘live’ with the content as metadata, wherever it travels on the internet,” Microsoft explained.
The second is a reader, which can be included in a browser extension and checks certificates and matches hashesto determine authenticity.
In its fight against deepfake, Microsoft has also partnered with the AI Foundation. Together, they will make the new tool available to organizations involved in the democratic process, including news outlets and political campaigns, through the Reality Defender 2020 .
The video controller will initially only be available through the initiative.
Microsoft will also collaborate with other companies to combat deepfakes, for testing, etc.
The University of Washington, deepfake detection company Sensity , and USA Today have also joined Microsoft's project.
“This process will help people distinguish misinformation from genuine facts and manage the risks posed by deepfakes and fakes,” Microsoft said. “Media literacy can allow all of us to think better about the context of media, while still appreciating satire and parody.”
Through the collaboration, there will be a joint statement encouraging people to pause and check to make sure information comes from a trusted news organizationbefore sharing or promoting it on social media, especially ahead of elections.
