Facebook
βWhen we find a piece of inappropriate content, technology can help us find all of its duplicate content and prevent it from spreading,β write Facebook head of security Antigone Davis and vice president of integrity Guy Rosen in a post. dedicated to the fourth annual Facebook Child Safety Hackathon. "For those who already use their own or other content matching technology, our technologies can provide another layer of protection, making security systems much more powerful."
Facebook claims that the two published algorithms - PDQ and TMK+PDQ - were designed to work with huge amounts of data and are based on existing models and implementations, including pHash, Microsoft's PhotoDNA, aHash and dHash. For example, the PDQ algorithm for photo comparison was inspired by pHash but developed entirely from scratch by developers at Facebook, while the TMK+PDQF video matching algorithm was created jointly by the Facebook AI Research Group and scientists from the University of Modena and Reggio Emilia in Italy.
Both algorithms analyze the files they are looking for thanks to short digital hashes, unique identifiers that help determine whether two files are the same or similar, even without the original image or video. Facebook notes that these hashes can be easily shared with other companies and non-profits, as well as industry partners through the Global Internet Forum on Counter Terrorism (GIFCT), so all companies with an interest in online security will also be able to remove content that Facebook has flagged as insecure. if it is uploaded to their services.
The development of PDQ and TMK+PDQ followed
In turn, Facebook CEO Mark Zuckerberg has long argued that AI will significantly reduce the number of abuses committed by millions of unscrupulous Facebook users in the near future. Indeed, published in May
Source: 3dnews.ru