Computer scientists at Boston University (BU) have created an algorithm that prevents deepfakes by adding an imperceptible filter to videos and photos before they are uploaded to the Internet.
If a deep neural network attempts to alter a protected image or video, the filter leaves the media unchanged, or completely distorted and rendered unusable as a deepfake.
The open source code has been made publicly available on GitHub.
Said BU's Sarah Adel Bargal, "We covered what we call 'white-box' attacks in our work, where the network and its parameters are known to the disruptor. A very important next step is to develop methods for 'black-box' attacks that can disrupt deepfake networks [in ways] inaccessible to the disruptor."
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found