From Extremetech:
"Guba.com, a video site that hosts user video, has developed "Johnny," a software tool to filter and block copyrighted content in cooperation with the Motion Picture Association of America. Guba which began offering downloadable Warner Bros. films in June, said its new "Johnny" tool will be made available to other hosting sites to block a list of copyrighted movies and TV shows that it received from the MPAA.
During Guba's early days, users came up with more and more sophisticated ways of manipulating images, such as adding text, and cropping the images. As they did so, the company began looking for ways to automate the detection process. The first image detection program turned images into MD5 hashes, a cryptographic tool also used to check for file integrity. The hash information treated the image as a graph, evaluating the image using a number of different attributes including image size, color variance, and other factors, Myers said, all mapped together.
Copyrighted images with hashes that matched other, unknown images were flagged for examination, and discarded if they violated the site's terms of service.
That program, known as "Johnny" works both on snippets of video, as well as trailers, half-hour episodes, or full-length movies, Myers said. The site receives a watch list of copyrighted movies and other materials from the MPAA, and cross-checks it against its hosted video, he said.
One of the keys that enables the process to work is the fact that Guba accepts video in a variety of formats, but transcodes it into its own file format. The automated process, which routes the video through the Guba servers, allows the company to create a hash during the process.
"Johnny" takes a series of snapshots, not of the video, but of the video hash, recorded every few minutes, and turned into a signature file. As new videos are uploaded, the files are cross-checked against the signature files."