Facebook is open-sourcing two technologies on GitHub to detect identical and nearly identical photos and videos in order to curb the spread of abusive content and share hashes — or digital fingerprints — of different types of harmful content, the company announced in its blog post. According to Facebook, the photo-matching algorithm, PDQ has been inspired from pHash although it was built from the ground up as a distinct algorithm with independent software implementation. The video-matching technology, TMK+PDQF, was developed together by Facebook’s Artificial Intelligence Research team and academics from the University of Modena and Reggio Emilia in Italy. How will the technology work? Both the technologies, PDQ and TMK+PDQF will store files as short digital hashes that can determine whether if two files are same or similar, even without using the original image or video. Also, digital hashes can be easily shared with other companies and non-profits. For example, when Facebook identifies terrorist propaganda on the platforms, it will remove it and hash it using a variety of techniques. These hashes will be then shared with industry partners, including smaller companies, through Global Internet Forum to Counter Terrorism (GIFCT) to allow them to take down the same content if it appears on their services. According to Facebook, both the technologies were designed to operate at a high scale, supporting video-frame-hashing and real-time applications. This technology is an addition to the company’s ongoing research in these areas, including research on new techniques to detect intentional adversarial manipulations of videos and…
