Facebook’s brand new, a high-performance algorithm, is unaware of its ever winning the war against fake accounts.
Facebook is opening up the tools that it uses to combat fake account development in its systems, and the company says that it has a new artificial intelligence method called Deep Entity Classification (DEC).
DEC is a machine learning algorithm that not only takes into consideration the actions of the suspect account, but also analyses all of the knowledge around it, including the conduct of the users and the communicating pages of the suspect account. Facebook says the total spam and fraud accounts have been decreased by 27%.
DEC has enabled Facebook to thwart over 6.5 billion fake accounts generated last year by scammers and other malicious actors. In reality, a vast majority of those accounts are in the creation process of an account, and even those that do move appear to be identified in automated systems on Facebook before an actual person ever records them.
Nevertheless, Facebook reports that around 5 percent of all 2,89 billion active monthly users are fake accounts under Facebook’s service terms and conditions. It usually means scammers, spammers, and people who try to phish vulnerable users or use other forms to access sensitive personal information for a kind of financial system or identity theft.
This is where DEC joins. The study of user behavior involves a complex and holistic approach that incorporates over 20,000 attributes per profile; for example, it takes account of an amicable action of a suspect and likely fraudulent account sent to by a buddy and not just to a suspicious account. The aim is to combat how malicious actors replicate real behavior. With time, Facebook says the advanced spammers would develop and grow in claiming that they are legitimate people, at least as the automated systems of Facebook perceive them. DEC can take advantage of this by exploring in more depth how to account identities communicate with the network.
This will be critical for Facebook as the US presidential elections are coming in 2020. Spam and spam accounts are just one facet of Facebook’s fake account epidemic. The organization has already acknowledged the use of social platforms to manipulate news stories, polling, and other critical elections through foreign operations from Iran, Russia, and elsewhere. And with time, these activities are becoming more complex.
Last year, Facebook and Twitter shut down an extensive network of fake accounts that used AI software to create profile photos in real terms. These images were not scraped, but they were created with neural networks that made it harder to flag as false. These are the types of tactics that will have Facebook working hard to keep a step ahead, and it knows that it has to continually rework its DEC strategy to ensure that it can remain effective against the evolving methods of spammers.