Deep Entity Classification

2019: Blocking of 6.6 billion fake accounts in Facebook

At the beginning of March, 2020 Facebook told about results of work of the Deep Entity Classification AI system allowing to distinguish and block fake accounts. Thanks to this technology only in 2019 removed 6.6 billion counterfeit accounts from social network.

Deep Entity Classification analyzes behavior of users by  a set of parameters and  collects  on  each about 20 thousand metrics. Besides, according to developers, a system independently studies and  automatically adapts to  new tools of spammers who constantly think out harder and harder methods to bypass artificial intelligence.

An AI system defined and deleted 6.6 billion counterfeit accounts in Facebook

The algorithm studies activity not only the account, but also users, group and pages with which it is connected. Such parts as the number of administrators in group into which the user, and distribution of friends by a floor enters are analyzed. A system connects different data points to find patterns which can indicate malicious activity.

Facebook claims that a system considers so many different parts that it can be considered much more difficult in comparison with traditional tools for detection of fraud using artificial intelligence. Usually such solutions consider a small number of factors therefore cybercriminals begin to understand how to deceive the computer.

After start of Deep Entity Classification the share of fake accounts in Facebook was reduced by 27% — from 5.3% of the general the number of active users to 3.8% by the beginning of March, 2020.

Despite it, Facebook doubts that it will be possible to get rid completely of counterfeit accounts as spammers quickly adapt to innovations and try to find the alternate paths allowing to use false accounts.^[1]

You See Also

• Counterfeit (fake) news (fake news)
• Fake (fake) reviews and comments

Notes