The AI systems automatically nab “policy violator” accounts.

Facebook (NASDAQ: FB) has finally achieved traction in its drive to stamp out fake accounts guilty of spreading spam, phishing links and malware. Its Deep Entity Classification, an AI-powered tool, will take down this menace. (TNW)

The social media giant has found itself under relentless regulatory glare for concerns on privacy, hate speech, fraud, and fake news. By its own estimates, 5% of active users are fraudulent. To combat this menace Facebook is making good on a promise Mark Zuckerberg made to US lawmakers in 2018. He said the social platform will deploy artificial intelligence (AI) in its fight against dubious accounts and content.

Facebook’s Deep Entity Classification

On Wednesday, FB revealed its AI modus operandi. It has deployed a machine learning model called Deep Entity Classification (DEC).

DEC, which has already detected hundreds of millions of account violators, is not anything like traditional methods of detecting fakes. These methods revolved around the user profile, IP addresses, number of likes and such. Instead, DEC looks deeper – into the page’s social interaction – the friends, groups, and pages with which it interacts. These three clusters of entities will give off their own signals. These signals are difficult to game or reproduce.

This multilayered approach allows the model to detect fake accounts. “This is defense-in-depth,” said Brad Shuttleworth, Facebook‘s Product Manager for Community Integrity.

Results showing progress

According to The Verge, Deep Entity Classification has helped Facebook stamp out more than 6.5 billion fake accounts that scammers and other malicious actors created or tried to create last year. Further, FB claims the estimated volume of spam and spam accounts has dropped by 27%.

However, FB may be running hard to stay in the same place – since about 5% of all monthly active accounts are fake – it will be a constant battle to weed out the bad actor accounts. But as its AI learns and gets better, FB may ultimately get the upper hand.

Related Story:  Artificial Intelligence: Facebook Bans Deepfake Videos… Advocates Worry