All the social networks are nowadays removing fake accounts using the machine learning model. The model is named as Deep Entity Classification (DEC). Facebook has already detected hundreds of millions of violating accounts using DEC.
DEC has no limitations of traditional machine learning models for detecting fake accounts like focusing profile pictures. DEC works very efficiently against highly suspicious behavior of Facebook accounts like posting thousands of same or alike photos. However, some people also fool this type of system by adapting the behavior of the system like how much friends it can have and how much time to be active.
Deep Entity Classification (DEC) can improve this by looking at the deeper features. DEC finds these by connecting a suspect account to all the friends, groups and pages with which it interacts.
Finding Fake Accounts
DEC will then investigate all of these connected entities. Friends of a fake account will give off signals through their ages, the groups they join, and the number of friends that they have. Facebook groups are checked by the members they attract and the admins that they have. Facebook pages can be evaluated by their number of admins.
After analyzing all these Facebook entities the model is enough trained to decide whether the account is Fake or Real.
Facebook Fake Accounts Report
Facebook covers, over 99.5% of the violating accounts they took action against were found before users had reported them in 2 years. Then also it is difficult to identify how much Fake accounts are being missed after this process. Facebook predicts that 5% of the new Facebook account created are fake monthly but they can’t be sure about the exact number.
DEC struggles to find political motivations. Dedicated teams are prepared by Facebook to investigate those types of attackers, such as an Information Unit that focuses on nation-state actors. DEC supports them for the Analysis of the data collected by the team.
Most Popular Articles: