Facebook engineers have developed an effective tool for detecting and blocking fake accounts. The system, which uses machine learning technology, blocked 6,6 billion fake accounts last year alone. It is noteworthy that this value does not take into account the "millions" of attempts to create fake accounts that are blocked daily.
The system is based on Deep Entity Classification technology, which uses machine learning to analyze not only active Facebook accounts, but also the behavior of each individual profile and its interaction with the rest of the community. During operation, the algorithm analyzes a large number of parameters related to individual accounts. DEC captures which groups a user joins, how many administrators and members are in these groups, when they were created, etc. It also analyzes the number of friend requests sent from one profile. One of the important points is that the system is able to automatically learn in the process of work, so it is constantly evolving as spammers adapt.
A Facebook spokesperson noted that the technology used to detect and block fake accounts helped reduce the number of accounts used by spammers by 27%. It is noted that currently the number of fake accounts on Facebook is approximately equal to 5% of the total number of accounts in use. Despite this, Facebook doubts that it will be possible to completely get rid of fake accounts, as spammers quickly adapt to innovations and try to find workarounds that allow the use of fake accounts.
Source: 3dnews.ru