Fake means something that is not original. Deepfake is a portmanteau of the two words – deep learning and fake. In a deepfake, deep learning algorithms are used to edit an image or video or audio in such a way that it is really difficult to point out that the edited image/video/audio is not an original one. Due to this nature of deepfake photos/videos, they have been in news lately especially because of fake porn and fake news. Also, deepfake videos of political leaders have been created wherein their faces are replaced or their lips are synchronised to speak a different audio track than the original one.
The image below shows an original and a deepfake. [source]
Deepfakes have taken the domain of image/video editing onto a whole new level wherein the entire process of image/video editing is automated via deep learning algorithms. In many cases manual intervention will not be needed at all.
Deepfake is a concern because it can be used to commit frauds by pretending to be someone else. Also, sharing of revenge porn leads to loss of privacy and dignity of a person. With advancement of algorithms, it is getting difficult to find out authenticity of audio/video/images which is also a cause of concern for law enforcement agencies.
For Wikipedia entry on Artificial Intelligence, click here.
For Wikipedia entry on Machine Learning, click here.
For Wikipedia entry on Deep Learning, click here.
For more posts on Artificial Intelligence, click here.
For more posts in The Cyber Cops project, click here.